The competition is real in pricing. Thanks for the Chinese open models, US big players have to cut their inference pricing. We've done a bunch of evals between the models, and Kimi K3 was the first one that actually could compete or be even better than Opus or Sol in our use cases, with a fraction of the price. All our developers use K3 as their programming model, and it now powers a big part of our systems instead of Opus and GPT. Surprisingly the new Sol pricing is quite similar to K3...
Now DeepSeek v4 Flash 0731 is eating Gemini's lunch, and suddenly we saw a price cut (the "introductory price") for 3.7. DeepSeek is of same quality or sometimes better than Gemini for text, Google knows it and they have to compete. Too bad it's too little and too late, it's still 4-5x more expensive in our evals.
And these models are not going away, nor their prices going up because of competition in the inference providers and due to the fact that you can buy/rent the hardware and run them in your own premises.
I shifted from DeepSeek v4 Flash 0731 to Gemini 3.7 flash on openrouter and price shoots up almost double with no visible change in outcome. So, today I reverted back.
I was struck by a video ad that Google released yesterday with testimonials by three developers about using Gemini 3.7 Flash [1]. The point they emphasize most is price, followed by latency. The marketing strategy definitely seems to be shifting.
Long-context agentic tasks and Rust engineering are our use cases where Kimi definitely is better than Sol. We can measure our own systems and the numbers say that Sol has no chance against K3 or Opus, and K3 is so so so much cheaper than Opus right now.
You cannot just look at the price tags for these models, you must eval and see the price per task. In our previous eval rounds Sol was more expensive than Opus (with its original price), took much longer, and provided worse results. Kimi does not have these issues, it's just as good as Opus with a smaller price tag.
China's 50 Cent Party being a real and noticeable thing (and the two biggest things they like to shill is open weight Chinese models and the futility of resisting a Taiwan invasion), I have to take things like this with a healthy dose of skepticism without corroborating data, since independent evals didn't show the price per task lead you're showing.
If there's independent data showing this feel free to share a link, I haven't seen it. DeepSWE has been most closely matching what I see in my own use.
And Fireworks did not yet. They are still under the limit of not feasible to self host... Let's see if other providers follow DeepSeek with their flash pricing.
After using Claude for a long time, I tested Sol 5.6 for the first time today. Love it, its an incredibly capable model and uses far fewer tokens/time thinking. Its what I imagine Fable would be if I haven't been downgraded on every conversation - even after completing the verification program. I think I may cancel my Claude subscription finally.
I switched from claude to gpt when 5.6 came out for the same reasons. I don't understand why so many people are still using claude when GPT and open source are so much better.
I think Fable's dominance is overstated. It definitely has the lead, but quantifying what that lead actually is is really hard. I'm using GPT 5.6 Sol to do some shit that I personally would consider "crazy" - low level undocumented hardware driver alchemy, reverse engineering highly obfuscated code, even a bit of screwing around with a rendering engine in Vulkan, really just about the most complex tasks I can get any model to do, and it does great. For the more advanced stuff, it definitely needs the effort bumped. But even with the effort bumped, the token usage really doesn't seem to skyrocket too badly until at least you hit xhigh and max, which really only seem to be necessary if you are doing genuine crazy stuff, so it's not that bad. I did similar stuff with Fable. In fact, I went directly from an Anthropic subscription with Fable to an OpenAI subscription with Sol, more or less, and it really felt pretty seamless. If anything, I was thrilled to realize how much I actually preferred Codex CLI, to the point where I started using it at work too.
Fable seems to be generally more impressive at outputting one-shot web apps. I'm not really saying that to try to downplay what Fable can do, it's just that if I compare the two, this is one of the few definitely noticeable areas that you can easily demonstrate. Obviously, one-shotting programs is much better as a demonstration of a model's capabilities than it is practically useful (not that it is useless, but hopefully my point is understood).
However, whatever Fable truly is better at, one thing I really like about GPT 5.6 Sol is even harder to quantify: taste. GPT 5.6 Sol outputs are still LLM outputs and they contain many things that people would probably consider "Claude-isms" for better or worse, but overall I really prefer the GPT 5.6 Sol output. I find it to be generally more tasteful. Hard to quantify, but when talking to people I've had enough people seemingly agree with me to convince me that it really is true.
I used Sol to extract the remaining decryption keys from the Super Mario Maker 2 (Switch) game files. Someone had previously extracted all the keys from the original release, but not any of the new ones from updates. Not only did it succeed, but it helped me understand the data sufficiently to add support for “Super World” rendering to my level viewer (which I made back in 2021), eg the little widget at the top of https://www.smm2-viewer.com/players/B16-306-GVG
I was very pleasantly surprised to find Sol wasn’t obstructive over what was clearly a very grey area endeavour.
Interesting. I've been wishing that old 'Stars!' game from the 90s would play easily on modern systems. I'd love it if we'd got to the point where I could point Codex at a folder with an ISO from my CD of the game and tell it to go reverse engineer it all for understanding of game mechanics, then go recreate it in a modern language capable of running cross-platform. Scarcely any need to improve on graphics, it could even be a PWA.
There's people that have tried to contact Jeff McBride and follow the IP trail but the IP is currently owned by a company that went defunct. Not sold, but no one is even bothering to register its LLC any more, it's simply dead.
I was having it look at creating a driver for some old scanner and it actively looked up exactly where that gray area for my country was wrt decompilation.
Fable is almost unusable for anything but super boring mainstream stuff. I was getting safeguard flagged so often I’ve significantly reduced my usage out of fear they will blacklist/ban me.
Some of the topics it’s flagged have been hard for me to understand what it seeing that can be remotely concerning in my requests.
I cancelled my Claude max subscription. Somehow every query I sent was flagged as bio or chem, even pure mathematics questions. Not going to waste money paying for a “max” subscription that won’t ever let me use the top tier model…
Sol is great and has never blocked a request, and generally gives great answers. Happily switched over to it now.
The dangerous part isn't that a model refuses extreme requests. It's when mundane requests become unpredictable enough that you stop trusting the model.
I've gotten flagged for asking questions about tokens and tensors. That makes me believe it's not about safety, it's about protecting their turf. I cancelled my subscription - same fear about getting flagged too much leading to a ban.
They said they also block usage of Claude models to build ML models.
Which is definitely protecting their turf, but also probably a little bit hiding their “RSI” abilities for competitive reasons. My theory is that a lot of “safety blocking” is actually WIP training of new business directions. Anthropic has started hiring biologists and has opened a preview of a “Claude code for bioinformatics”. I’m guessing they’re tweaking their bioinformatics market play, and block “bio safety” requests so competitors can’t learn about their training.
I feel like those examples are considered difficult because they're niche topics, but aren't actually all that difficult in a general sense. What I consider truly difficult are things like taking a ticket and implementing it in a preexisting codebase, using a clean and reasonable design that fits the existing style and makes sense to a human, and avoids the footguns I learned by working with the codebase for over a day.
If you said this in 2025 I would've 100% understood, but to be honest getting AI models to do a pretty good job on day-to-day ticket work has become so boring that we don't even bother using the top tier models and higher effort slots for that anymore. I personally wind up tweaking the results a lot and recursively having fresh agents review the diff, but that's just because I'm picky; in a lot of cases the first diff is actually pretty damn decent.
Compared to what I am doing at home experimentally, I feel like day-to-day work is absolutely nothing. Not only am I also working with existing codebases in my experimental prototyping, but I am also doing things vastly more complex with vastly harder constraints.
I use the AGENTS.md to show it how i want the code to look like. Something like "when implementing hooks adhere to the guidelines in docs/react-hooks.md". And then react-hooks describes your heuristics and what you consider best practices. There is a clear difference in code quality for me when using codex with a well crafted AGENTS.md vs. without one, you can run the experiment yourself pretty easily. As I mentioned in another comment, I think Claude poisoned users to stop relying on their Claude.md files and new codex users might be surprised at how well it adheres to guidelines.
FYI I run it consistently in xhigh regardless of difficulty of the task at hand. I remember high being very fast, but I'd rather wait a bit more and get better output. AIs are insanely fast compared to me anyway, even on xhigh. Consumes more usage, but even at 100 EUR/m I don't hit limits.
After hitting the session limit on my company's plan so many times with Claude when I was using it, I mostly keep Codex on "high" rather than "xhigh" as a way to leave the tokens for my more ambitious coworkers. It's possible that having it higher might end up with better output, but so far at least I've yet to see a way to get any model to do 100% of what I need up front without any need for me to make changes that end up being more tedious to do via interaction than by hand, and it doesn't feel worth spending a bunch more tokens trying to figure out how to better communicate to it up front how the dominoes get set up so they fall in place properly the next time.
To be fair, I actually do run xhigh as my default. However, for the first time in my experience of trying and using LLMs, with Sol.. sometimes I feel confident enough to set the effort level to "Low". I just had Sol prototype some AWS stuff on low earlier. Great result, did exactly what I wanted.
Initially I was planning heavily around context limits, but I've learned to just ignore it completely. Compaction is seamless for me. If details are lost in compaction, the model just re-reads what's needed. My conclusion is that at least for Sol, the summaries (which I've never seen) must be amazing. Every now and then a detail gets lost and I have to repeat it. I don't think there is performance degration, because the model is smart enough to re-read relevant files as needed.
And Mai-Code-1.1-Flash seems like a really good cooperative player to GPT 5.6 Sol. You get Sol to help you make a detailed plan, and Mai codes it up and you can get pretty decent code out the other end without too many tokens if you are careful.
things that are alchemical are rarely alchemy. That is to say things are very fiddly but stick a room of monkeys on typewriters, a schizophrenic developer with HolyC and adderall or an LLM, persistence is the key to many of these things like drivers, extracting keys from vintage security domains, etc. Dropping into xdd to a human is a chore, not for an LLM.
Although I am not exactly sure what you mean, I am not really claiming it is doing anything I couldn't do - but yes, it does so with much less effort. For example, I can have it set up probes and tracing on Linux that I personally would have to consult documentation to do. It might not even have to consult the documentation due to having the information on-tap, but even if it does, it's nothing that would cause it any fatigue, it's just going to keep moving forward in a loop until it is satisfied that it meets the criteria. I could've done all of this alone - I really could have. I just would not have. Being able to do something 10 times faster or with 10 times less effort is, in some senses, sometimes more impactful than being able to do entirely new things you couldn't do before.
This is a really good point that I definitely failed to grasp when first hearing about these tools. At least for me, the best way to use these tools is as a way to free myself from having to spend time thinking about the things that aren't worthwhile so I can focus on the things that truly are. I've had times in my life spending hours reading documentation and googling random things to try to tease out the correct sequence of commands or the exact right shape of an API to be able to make things work to know that it doesn't make me more productive to do that myself rather than point an LLM at the thing and let it spit out the answer after a few minutes. Meanwhile, I can spend that time thinking about what comes next, or what the correct way to take that one-off output and abstract it to something that can be used meaningfully in more flexible ways.
The only obvious objection I can think of to this line of thinking (at least from a technical perspective) is "how does someone build up the knowledge to be able to use a tool effectively in that way if not by doing things by hand at first?" The honest answer that is "I don't know, but that's also pretty much exactly the type of thing my employers have never been paying me to solve in the first place". Even just a decade into my career, there have already been plenty of times in my career I've struggle to convince people that we should do stuff in a way that won't bite us in the ass a month or two down the line, and in the times I've managed to succeed, it's usually only by putting in more of my own time and effort to make the initial investment seem more palatable. Luckily right now I'm not in one of those times when I'm having to go full throttle to keep the lights on a few months from now, but I don't have enough fuel in reserves to work on a plan for when we need to build a new rocket in another ten years. Maybe ask me next month.
Sol is way too eager to hone in on small details and ends up with massive over-engineering. Fable does it too - to be fair - but noticeably less.
After extensively using both on Max 20x plans, I've concluded that Fable is better for problem solving and coding, whereas Sol 5.6 Ultra shines in debugging specific issues: tackle a problem with Fable then leverage Sol to clean up, double check, or fix specific issues.
Fable (imo) had the edge on the $200 plan, but after this 50% reduction I'd say Codex is better value by far and there's no contest.
---
Using Fable as the orchestrator and delegating tasks to Sol 5.6 Ultra via the codex plugin in Claude Code yielded good results, but still there was a lot more over-engineering (thus time and tokens spent) than Fable by itself would've done.
Both models suffer from doing-too-much. But both models are fundamentally really smart and knowledgeable. I think it's really close and pricing cuts really spice things up for us consumers! Sol is a clear winner in the value department and the $100 plan is enticing!
---
*Claude Code usage is reducing by 33% in 2 days, Wednesday August 19... cmon anthropic: clau.de/cc-50-promo
Yeah, I very much agree on this. I think Sol and Fable code quality is on par. Maybe Fable is just a tiny bit better, but Sol compensates with its ability to work through things, while Fable, in my experience, generally tends to avoid solving problems that require many LOC.
However, I think these are very different models in terms of orchestration. Long-horizon tasks are way more predictable with Fable. It just doesn't lose track of details. Thus I ended up building a small wrapper around Pi (where I run Sol) so that CC can delegate via background tasks, automatically wait for completion, and do what was one of the most effective parts - steer Sol toward simplicity, getting Sol out of code-review infinite loops (Pi calls for Codex review to ship better, but generally gets stuck on P2 and results in vastly overengineered work).
One of the worst experiments was enforcing coverage at 100%. Only Sol, with an enormous amount of code and significant pushback (on architecture decisions) to Fable, was able to reach it. It made me think this is somehow related to overengineering in general, so that instructions on acceptance criteria in claude.md plus proper DX (e.g., Lefthook) actually led to okay results. It mostly helped that responsibilities were clearly split: Fable designs architecture, Sol handles coding and debugging.
It's great, don't get me wrong, but so is Fable. I'm just comparing the long-horizon task performance between the two at the same or similar effort levels.
Given the 50% discount on Sol and how smart it is, yeah it's unprecedented value. If you only want to use low effort, there's a clear winner here on value and it's not even close!
*The "Max" I referred to was the plan tier, not the effort level btw
For small tasks, you can just use something like low or medium effort and it can usually avoid mistakes; after all, the model will test the code anyways and can do some baseline level of iterating.
In regards to cost, we need to acknowledge how generous OpenAI was in the last couple months with Codex usage credits (no weekly limits) and usage resets. It afforded me many a dive with Codex! Yes it uses more tokens, but sometimes it's worth it -- just depends on what you're working on.
Finally, Ultra(code) isn't that bad when it comes to cached tokens. I think folks overstate the general token usage of ultra effort on both providers.
---
Both models are great at green-fielding a project when given detailed specs.
Both models overthink too liberally (imo) during these larger multi-shots. Sol overthinks more than Fable.
Both models are really smart and perform great for general knowledge and regular coding tasks.
5.6 Sol is a joy to use for "daily chat" as well. Compared to earlier OpenAI models it catches and corrects its mistakes very reliably. It also seems way smarter in tuning its replies to areas I am more/less knowledgeable about (i.e. when I ask it a law question, it assumes I know as much as a toddler which is true, but on political topics it more easily throws around terminology) and including analogies. On medium thinking, it's a very good compromise between speed and quality.
Fable feels less cumbersome to work with, but it is SO DAMN ANNOYING with the refusals that I'm leaning more and more on Sol, and very much looking forward to GPT6. Just seems like Anthropic is trying their hardest to ruin their reputation and user experience.
Yeah I feel like I'm living in a different dimension than these people. I wonder what they're working on. I've literally never had it refuse everything and I max out my 20x plan every week
Can I ask which country you're in? I have a theory that the safeguards differ depending on the user's country.
I'm in Australia, and Fable downgrades to Opus when testing for bugs in memory in a legacy C code base. If Fable starts taking initiative and writes a test case that involves writing to a null pointer, that's the end of the conversation.
As one random example, today I had it hit a refusal loop when adding a country selector dropdown to a form, presumably because it contained a “bad” country name? I hit refusals at least 2-3 times per day, sometimes many more. The worst part is it is often right in the middle of a multi-stage task, so the only option is really to switch to Opus 5 and let it defecate its absurdly verbose comments all over the rest of the edits in the turn and hope it doesn’t go on one of its tangents, then have Sol do damage control. Oh and I got approved for their “cyber verification program” blessing, which comically does absolutely nothing for Fable.
I too switched to OpenAI after I got sick of Anthropic's constant "safety" downgrades. Sol is definitely a breath of fresh air.
> even after completing the verification program
Was it easy to complete it?
I ended up in some weird state where I can't even attempt the verification at all. Opened the Persona tab once, closed it and then it never opened ever again. It says a verification precheck failed.
Even without TAC, Sol doesn't seem to get blocked very often. Fable would downgrade to Opus if I looked at it wrong.
I cancelled my subscription recently and moved to Sol. So far - it has been a great experience. The only aspect where Fable/Claude is better I feel is doing some research from the web and summarising the facts.
Sol is my daily driver but there are still times I reach for Fable when Sol doesn’t cut it. Just yesterday for example, I was trying to build a self-modifying hot-reloaded agent harness in Elixir for fun and Sol just kept doing silly things like thin wrappers and unnecessary abstractions. Fable handled the task elegantly. Sol is really good as a reviewer for finding bugs due to its thoroughness however.
It's the complete opposite for me. The model might be the worst model I have ever used when compared to other models in the class. You just can't get it not to just write the most enterprise over complex over engineered solutions for every little thing you ask it to do.
It the first model to actually make me pissed off to use AI. I absolutely hate the model so much.
I don't even want to see the codebases this model is fucking up.
It might just be good at finding bugs that about it. That all I would ever use it for just because it works harder than Claude models.
Used claude since 7/2025. Switched to codex after fable got blocked. It was still 5.5 but I knew they had to come up with something. As soon as I switched, wow. It wasn't super intelligent, but it was stable. Every day it was the same performance. This consistency is definitely worth paying for.
Yeah at this point claude is overrated, overly expensive, weird writing style (elliptical), and the worst part is the aggressive guardrails that even normal convos get interrupted, meanwhile openAI is still I would say at the normal balance, if you ask something too obvious or direct it will stop you other than that, it work flawlessly, plus, I have yet to hit the limit despite heavily using it these past weeks.
sol is much better imho than Fable but i can understand if they will perform wildly different for different people with different levels of expertise aswell as different needs. I dislike fable myself it doesnt really work for me.
Sol also doesnt _really_ work but it sort of tricks me into thinking it does more convincingly :p.
cancelled my subscriptions few days ago. (was on 100$ ones, not sure if there is diff in quality for higher tiers or not.. there might be that too).
what i hate the most is that they will make any obvious mistake you do not tell them to avoid. then on the next plan to fix it, your token limit is hit at step 4/5 -_-. Both models seem incredibly good at that mostly...
for tasks outside of coding and program design i do find them quite useful. like devops crap. maybe because i hate that, i like their help there more.
The final straw for Claude was its refusal to give me a list of the most recent rapes reported by the BBC and basic information about them (location, date, names, just things reported in mainstream media). It outright REFUSED to complete this task.
I will not be told what I can and can't do by AI and I will no longer be supporting American companies run by despicable people. GPT only gets my money right now because its so fast and cheap but I'll be back to Chinese models in no time.
I have a "strategy / life-coach" project, and was surprised at how much better Sol is than Fable on it, as I've found Fable to have the edge for most things for me so far. But Sol: questions were better, insight was better, it got the brief better.
Sol has held stuff for a while to do the same sort of hazard checks I assume Fable is doing, but it always releases them. I think that's the better way to handle it rather than preventing me from seeing how far I can get generating schematics to use in Minecraft. Currently: a mostly normal voxel house.
That was an issue at least a year ago. I had signed up for a claude account on my iPhone and then wanted to sign in on my laptop but nope, not possible. Insane they still haven't fixed it.
Can somebody at Anthropic tag claude in slack or whatever goofy shit you do and ask it to add Apple OAuth to your website? Clearly humans aren't testing it.
Always has been. My prediction is that both OpenAI and Claude will go bust unless they deliver a killer product. And unlike scrappy startups, they have a pretty serious deadline because creditors will come a-knockin'.
There's little to no functional difference between Kimi, Qwen, Sol, Opus, etc. All flagship models are within like 1-5% of each other and the real moat will be what's always been the hard part: making a good product.
> All flagship models are within like 1-5% of each other
Don't know about that.
I'm using code review of my lone lisp project as a benchmark. It's a massive parallel code review where a coordinator cuts up the codebase into sections and dispatches agents to consider each part from different perspectives like quality, maintainability, consistency, correctness, rigor, etc.
Ran a complete Fable/max code review. Took over a month on a subscription. Now I've switched to OpenAI and am repeating the exact same review with Sol/max.
It's still not done yet but preliminary findings suggest Sol can only reproduce 70-90% of Fable's findings. So I think these models aren't as close as we've been led to believe.
The problem is that most of the volume doesn't come from proprietary products, it comes from API use which has no stickiness.
Claude already has a killer product (claude.ai/chat is a Swiss army knife) but just relying on people typing stuff into chat is not enough to sustain the company.
The other strategy is entrenching yourself as the LLM of choice into existing products (like ChatGPT is on Apple products).
There is a massive difference even between Opus and Fable, same provider, before various harnesses and other optimizations come into play. Don't be deceived by rankings and benchmarks, try for yourself.
I don’t think that’s true. OpenAI docs don’t have this price change. I assume they’d be the source for this post if it was true. The banner on OpenRouter for me says Gemini 3.7 discounted for a limited time, but if I click through that I get to this page: https://openrouter.ai/models?discount=true
That shows a bunch of models, including Sol, with a discount. None of them say how long it’s for, but I’d assume in all their cases it’s for a limited time as the banner said, and only on OpenRouter.
As I understand OAI is offering discount only for users using the model via OR. Not sure why. Maybe they want OR users to try the model and switch to OAI subscription or something.
It might also be worth noting that the discount is only if OpenAI itself is used as the provider via OpenRouter. The discount does not exist for Azure or Amazon hosted Sol.
I posted elsewhere, but the Azure uptime & performance for Sol is truly dire. OpenAI is offering 5x faster latency, 4x faster tokens generation, and vastly better uptime (Azure US has only 87% uptime), all for 50% of the price now. I assume the pricing is to compete with other shiny new models (Grok, Qwen etc), but it might also be to cut-off a truly poorly performing Microsoft hosting experience.
A surprising amount of companies sell exact the same product through different channels for different prices. A good example is Apple.
Multiple times a year, retailers here in Australia have co-ordinated sales on Apple products. Apple.com or their retail stores don't have these sales.
But they're clearly Apple-funded when competing retailers launch the same sales on the same days; and the margins aren't enough for retailers to take a loss.
Or it's "normal" market segmentation. OpenRouter users are more price sensitive in general, also a lot of enterprise users who can't switch easily are using the official API (or Bedrock or Azure) and you want to squeeze them as much as you can.
Seems reasonably clear to me? Potentially bringing in more customers who use OpenRouter for trying out all the models with rapid switching, enticing them to use Sol. And anyone being routed on price will immediately be switched to OpenAI's servers instead.
It also seems to be providing a vastly better user experience - Azure has less than 99% uptime (Azure USA only has 87% uptime), latency of 20 - 30 seconds, and a mere 8 tokens per second. OpenAI is offering 32 tokens per second (4x faster), 4 seconds latency (5x faster), and all for half the price of what Microsoft is charging for a vastly inferior experience.
Luna saw a huge jump after the price cut and is one of the more competitive models at the new price on openrouter.
Maybe they want to see how much market they can grab with Sol?
This might help but there are already cheaper models with Sol's intelligence more or less, the most notable being Grok 4.6 at $6/m which makes it a tougher sell
It's really only between Anthropic and OpenAI for many of my use cases, since I have a Zero Data Retention agreement with both. I'm not trusting random inference providers and especially not Elmo with sensitive data.
or Tinfoil [0]? They serve open models with container integrity attested by Nvidia/AMD enclaves. Every cloud provider offers this of course, but not usually in a way that can be shared between distrusting users for economical inference. It still relies on the open-source containers being secure, and there's probably hardware sidechannels and stuff, but personally (ie privacy not liability) I trust it more than a contract
DigitalOcean does inference, has good uptime and offers ZDR. They are not an AI first/inference first company. They have been serving cloud products for almost 2 decades.
Batch API is a bit harder, not many models/providers support Batch. It primarily is only the Gemini/ChatGPT/Claude models that do. DigitalOcean does support a 50% discount on Batch API via them directly, not listed on OpenRouter.
I have switched to Chagpt sub now after only using Claude for coding. You get more value for your money and feels like codex has reached Claude code performance in coding (the reason for using Claude) regular plus account allows you to have access to their most powerful model, image generation and asking questions is better because you can use sol but in instant mode and it feels smarter and faster.
And finally codex usage limits are better than the Claude daily 5h limit.
And codex feels faster although Claude code had more features
I switched as I felt Codex was on a par with Opus, but the chat responses from Sol are just more intelligible than the word soup I've been getting from Opus. I wonder if Opus could be prompted to respond in simpler prose via agents.md
Ah, so I am not the only one struggling with deciphering Opus writing style. At times I feel dumb as a rock because I read the same passage like 5 times and I still don't get it.
> I wonder if Opus could be prompted to respond in simpler prose via agents.md
God knows I've tried. I've got a variant of the ASD-STE100 trick which does the job, mostly, at the start … but get to about 100k of context and it goes out the window.
The model's personality is too strong for simple suggestion, alas.
I have also noticed that I am increasingly struggling to read what LLMs are writing, finding it incomprehensible half the time.
I stole Matt Pocock's line of "When reporting information to me, be extremely concise and sacrifice grammar for the sake of concision." for the agents.md
It makes it a little better. I also specify to use https://github.com/AminBlg/SimpleEnglish/ for all writing it does including code comments. It all feels like a bandaids but that seems to be the best we can do right now.
I used over a billion tokens per day of gpt-5.6 sol xhigh starting last Wednesday through Sunday before reaching my reset limit. The $200 pro plan is still the best deal.
Pretty basic. The codex app with one conversation per project and several running simultaneously all hours. I’m going for max caching that way and it never gets lost even with compaction somehow. Each has a plan with milestones to keep up to date and a thin agents file. I check in on them in the Remote app. Use case is protocol and control reverse engineering of audio hardware. I think they must be identifying the heavy use agent sessions and cranking up their cache lives so it’s not a big deal for them.
I'd like to see a benchmark on this specific topic: Reverse engineer the hardware protocol from a driver, or just migrate a driver from one OS to another.
Ultra mode spins up many sub-agents. On a particularly challenging task, I’ve had as many as 29 agents working at one time.
Also if you don’t specify, most end up being the same as the parent model which is pretty wasteful.
I engineered a skill that spins up Terra High agents for most sub-agents, resorting to Sol Medium for technical research and Luna High for code/in-project research tasks.
On a slightly different topic, Luna Max is incredibly capable and doesn’t use as much quota (Luna tokens are dirt cheap).
Most importantly, are you seeing a return on investment for time and ultimate outcome?
No one can judge the enjoyment, learning, and hobby aspects. Just wondering if there is an end goal for that much overall expenditure (time, money, energy, etc.)
I absolutely see a return. I do not pay for this (we have a corporate gateway), for the price of a junior developer I can get 3-4-5 senior developer's work done. it's insane value.
Some people just do crazy stuff. For example this now ex yc guy who said he has agents constantly scanning Sf govt apis and forming dashboards just because
Yeah I am, building large software with a vision - requirements - architecture - plan - code workflow. One Claude max account is enough to work on one, maybe two of those at a time (call it 15B tokens/month per project)
I’m exclusively using ultra and I run out in 3-4 days consistently. Those resets are great but I’ve noticed they like to cluster them at the start of the cycle, would be better if they spaced them out more.
A billion tokens per day?? Plausible estimates put the energy use at about 0.001 Wh/token, which means you're using 1000 kWh/day in electricity, just to generate slop. That's about the same as 50-100 houses. 300kg of CO2 per day - roughly the same as flying from London to New York every three days.
I think on average AI energy usage is not as big a deal as everyone is panicking about, but your usage is truly absurd and I don't know how you can live with that. It's immoral.
I can't speak for that guy, but I'm a physicist and work in clean energy... So it's not too hard! That said, I usually am closer to 10M on days I do heavy coding, so not nearly that bad.
As a company... but that includes things like research costs, model training etc. to determine if they're selling electricity at a loss you should look at inference costs bc that's the "thing" they're selling
I can’t sign up for that. I tried authorizing Codex a couple days ago. For some reason, their system says my phone number has been used for verification 3 times even though it definitely has not. I’ve had this phone number for over 20 years. OpenAI support is useless. They just keep repeating the policy without actually helping me.
Use TextVerified, load up like $5 of credit and OAI verification is like $1.00. Then when your account is made, ensure 2FA/passkey is setup then you don't need to worry about the phone number.
Yes I filed a support ticket with them and explained that their system is broken and they just did not care. I explained how it was impossible for me to use it 3 times already as I've only made 2 chatgpt accounts EVER, and only recalling entering my phone number for one of the two chatgpt accounts. I told them that this issue locked me out of codex and chatgpt for work and they weren't willing to do anything about it. Totally useless support.
I ended up borrowing my gf's phone number just so I could get access for work. Ridiculous
Historically those are less useful because some of the verification systems require a real phone number and that your name is associated with the account, depending on what and how they verify. It's annoying, I use a google voice number as my primary, and it often gets rejected.
If they can cut the price of Sol by 50% and the price of Luna by 80%, then the original price might have carried a massive operating margin. They might still be serving the models at a profit after these price cuts, but we will never know.
I don’t think there’s a real answer for this. Margin depends on whatever number the accounting department wants to make up.
Do you include research and training costs? Of all models or only the ones being served? What percent of the R&D budget do you allocate to inference? What about data center capacity? Do you count future commitments? All the circular financing deals? Do you count employee equity grants as costs? At what valuation?
We also have another solution for "whatever accounting decides": generally accepted accounting practices. It's far from perfect, but GAAP figures are what you should be looking at; not "adjusted GAAP" or whatever invention.
It is a well-accepted fact new shoes make you faster. Current science suggests it is due to the lighter weight from lack of dirt, though there is a competing theory which says it's an optical illusion due to the fact the pure white streaks resemble speedforce.
This is the opening salvos of an all out token price war.
With models a commodity at this point there isn’t much leverage for the big labs to keep their pricing anywhere near where it’s at. And that’s at the worst possible time as they need to be dramatically raising prices to have a viable business model.
Expect pricing to rapidly fall towards the underlying cost of compute and as players get really desperate we’ll likely see inference at less than the cost of compute as the market starts to rationalize and squeeze out weaker players who’s only play left will be to be the cheapest option in town.
The AI bubble is just waiting for the first player to scream mercy and cut capex as they simply can’t afford to throw more cash on the burning pile. That will be the trigger that implodes this bubble.
Since it's only discounted on the standard "OpenAI," non-ZDR route (old pricing on Azure), I'm guessing a lot of users won't see this benefit? Since a lot of users enable a global "ZDR-only" toggle on OR
That suggests this is being funded by OpenRouter, and there's no indication this is the case (and I doubt OpenRouter can afford it; why would they anyway).
I would in such a scenario expect the GPUs to be dumped to industrial breakers who would send them to China for refurbishment and repackaging before being sold again on Amazon, AliExpress, and Taobao as last gen gaming cards from weird brands and specs.
This is what happened after the great crypto GPU dumping.
The e-waste recyclers are pretty low on the pecking order, as the creditors will be first to strip these places for assets as Leopold Aschenbrenner discovered. =3
Yeah, I ment it as a joke - I agree with you. Watched the Gamers Nexus GPU investigation recently, where they were shown how a chinese soldering shop can transplant GPU chips to a new board, including memory chip reuse.
Hopefully we can look forward to all that useless datacenter AI crap gets repurposed in a similar manner into something actually useful for users.
I like to see this. I still prefer Fable (marginally) but my last big task was 100% Codex using Sol max (re-sizing my AWS infrastructure using CDK) and it did a very good job. No complaints, I could use this model happily to do what I need to get done.
If this nudges Anthropic to give me more Fable usage, that's even better.
> my last big task was 100% Codex using Sol max (re-sizing my AWS infrastructure using CDK)
Fwiw, you could do this with any small or medium model, and it's easier with the aws-docs mcp. AWS is pretty stable, well documented, and programmatic, so most AI can figure out what it needs pretty quick
OpenRouter doesn't do sales, they charge a premium, which is a flat 5.5% taken out of your credits. If you see something cheap on OpenRouter, it's because that one provider lowered its price. (Actually, correction, they will take 0.5% off their fee if you allow them to train your content)
Another thing some people don't notice is flex pricing, which is way lower than default pricing, for slightly worse latency and reliability. Depends on the provider and model
My bet is that OpenRouter began steering GPT-5.6-sol users towards flex tier, which is already 50% off.
So this isn’t really a price cut. As to why, lots of possible reasons. Perhaps an agreement with OpenAI to help them drive up more diverse traffic priorities.
I'm not having that experience. So far each major model update has been at least slightly better than the last, in ways I've found useful. Can't say it's perfect, or able to do exactly what I want without a decent amount of instruction/implementation/docs, but it's been useful enough to keep paying for it.
Oh no the models are absolutely getting better, I'm just amazed that only 6 months ago I was using gpt-5.3-codex, and now I can use gpt-5.6-luna for similar results at like 1/15th the cost. Now 5.6-sol is being slashed by 50%? Amazing.
You can train a LLM to inverse summarised thinking into thinking text. It’s not perfect, but it gets you maybe 80% of the quality with proper techniques.
FWIW, there’s not that much value protected here anyway IMHO, and even raw thinking text can lie (as shown by Anthropic’s amazing research), so for legitimate interpretability research it’s limited.
Scaling frontier performance hasn’t been SFT-bounded for a while now; it’s now basically how much you can scale RL rollouts.
I saw this for Luna and then looked at the uptime and it said 85%. My interpretation is that this is just a gimmick where they serve the OpenAI flex tier at the same discount OpenAI provides for flex and then fall back to azure
Has anyone had mixed experience running Ultra with and without /goal?
I come back to it after 8 hours to find it got stuck navel gazing imagined and Byzantine errors.
In my experience, "Sol" stands for "Stupid overengineering LLM". I’ve tried it at low/medium/high/xhigh effort levels and after a while I always end up to regretting my switch from Opus/Fable.
Reading the comments in this thread, i honestly dont get it. 5.6-sol has felt like a regression in capability. In fact, every model since 5.3-codex has been a regression from OpenAI. I just find 5.6-Sol over engineers problems, takes absolutely ages to solve basic problems....
At this point, I'm considering going back to cursor over codex due to the ability to get more control over what model I use since there is clearly a heap of user preference and having frontier providers constantly shift the goal post with "State of the Art" is complete non-sense.
It depends on what effort you're using etc. As an example [1] of what codex is capable of, here's hugo (written in golang) ported to TypeScript - and then a TypeScript to Rust transpiler which converts arbitrary TypeScript into Rust.
The TypeScript code which was transpiled into Rust (and is compatible with most hugo templates) runs faster than the original hugo.
I find all effort levels of sol are the same in terms of amount of hallucinated unnecessary changes. Luna is much better all round on xhigh but my point still stands, every release of these new models is not an upgrade, its re-learning how to work with it.
Its like rehiring an employee every few months then training them up. Its honestly tiring and cant stay like this.
That's not been my experience. My prompting methods haven't changed much between recent GPT releases. I do put a lot of effort into building tooling and tests around a project, so the LLM output is converging around it.
I mean it added additional changes when it doesnt need to. Its basically hallucinating changes it thinks it needs to make regardless of effort levels i try.
These threads seem to have become exceedingly vibes-based.
Yes, something may now be cheaper or more expensive or whatever, but there is no way to objectively measure quality (except for "trust me bro" benchmarks). So the discourse is people saying that for them, this or that model was better - which is a very low value data point.
Yes, programming language discussions can be vibes-based and therefore low value too, but sometimes the more concrete aspects of the languages at hand, e.g. language features and tradeoffs are discussed. That is something that I'm not seeing in equivalent AI discussions.. there is a lot of how a particular AI model "feels" to interact with.
No, this is OpenAI doing the discount, not Openrouter by themselves. OpenAI is crushing it with their 5.6 models, and they probably decided there was no better time to grab as much market share as possible.
I don’t understand this at all. They have never been profitable yet. How is this helping them? When it be more likely the case that not enough, people are using it as the prices they established already? So now they have to lower the prices?
You lower prices for marketshare. Fable became a mythological model to leadership because they were the first story of ai escaping and hacking another company. The it's so dangerous the public can't use it narrative is sticky so OpenAI is showing off its model so as many eyeballs as possible. We're in the samples in the supermarket phase.
Not sure as OpenAI models (Sol, Luna,..) are also discounted on the Vercel AI Gateway rn. My bet is on OpenAI trying to drive more enterprise customers to their models through API.
It shocks me how little people seem to care that they are supporting an evil Zionist lizard man who molested his sister and is happy supporting trump. Doesn’t even come up in the conversation here. I don’t really care if sol is a bit better, I still make decisions on more than that.
Is the HN community just too online and sucked in to the musk mind manipulation vortex? Or what is going on? Why does nobody seem to care?
I stopped using superpowers because it wanted to turn every tiny bug fix into a $37MM DOD project. I got effective results but it took ages. I may try again - I need to find a good way to run different profiles in my harness so I can easily shut it off. The default planning workflow in OMP is pretty good though.
I agree Luna is great for task execution, either as a sub-agent with Sol planning and coordinating or if the task is well defined and straightforward, but there are lots of models now that you can say that about.
You would probably get better results with Luna for the real simple tasks, or Sol with low thinking effort.
I find that I get exactly the effort that I asked for, which is pretty nice. The other side of that coin is that these are the least lazy models I’ve used so far. They will go on elaborate tangents to complete the task when I want them to.
I've found it's worse for simple tasks too, and I have to give it stricter guidelines, and sometimes it doesn't follow the same patterns I've grown to expect. I've found using 5.6 (sol) is good for diagnosing issues though, especially in terms of optimization of some given path
Yep. I have not yet had a single good experience with Sol or the 5.6 models on a variety of harnesses and configurations. It overthinks, overcomplicates and often makes my code into an unmaintainable sludge. It'll usually take 5+ turns of steering to get it in the right direction.
It's your responsibility to set an appropriate level of Thinking. For simple tasks, I use the instant model. As an approximation, the choice is proportional to the amount of time I want it spending on the task. Also, you can always ask it to respond succinctly.
I’ve used Claude exclusively for the past few months
Was excited when Sol came out a few weeks ago and loaded it up
I made the mistake of treating it as if it were Claude - I’d assumed they were close enough in ability and treated them that way
Well, turns out my instruction sets for Claude are 100% too complicated for Sol
Sol made the stupidest assumptions, constantly did things that it wasn’t asked to do and always approached code in what I considered a weird way - I had redo a lot of my prompts to get it anywhere close
Now, did it do good work?
Yes, on occasion. But with LLMs and coding, consistency is the name of the game. Constantly having to correct the LLM and constantly feeling paranoid that it won’t listen makes for an exhausting session
Maybe if you “came up” in the codex world you’re more fluent with it, but sticking with Claude for now
The competition is real in pricing. Thanks for the Chinese open models, US big players have to cut their inference pricing. We've done a bunch of evals between the models, and Kimi K3 was the first one that actually could compete or be even better than Opus or Sol in our use cases, with a fraction of the price. All our developers use K3 as their programming model, and it now powers a big part of our systems instead of Opus and GPT. Surprisingly the new Sol pricing is quite similar to K3...
Now DeepSeek v4 Flash 0731 is eating Gemini's lunch, and suddenly we saw a price cut (the "introductory price") for 3.7. DeepSeek is of same quality or sometimes better than Gemini for text, Google knows it and they have to compete. Too bad it's too little and too late, it's still 4-5x more expensive in our evals.
And these models are not going away, nor their prices going up because of competition in the inference providers and due to the fact that you can buy/rent the hardware and run them in your own premises.
I shifted from DeepSeek v4 Flash 0731 to Gemini 3.7 flash on openrouter and price shoots up almost double with no visible change in outcome. So, today I reverted back.
I was struck by a video ad that Google released yesterday with testimonials by three developers about using Gemini 3.7 Flash [1]. The point they emphasize most is price, followed by latency. The marketing strategy definitely seems to be shifting.
[1] https://youtu.be/kacf2bib-X0
> Opus or Sol in our use cases, with a fraction of the price.
I assume it's highly use case dependent, though?
Even before the price cut seems like Sol was price competitive with Kimi
https://artificialanalysis.ai/models?models=gpt-5-6-sol-xhig...
And now it should be considerably cheaper
Long-context agentic tasks and Rust engineering are our use cases where Kimi definitely is better than Sol. We can measure our own systems and the numbers say that Sol has no chance against K3 or Opus, and K3 is so so so much cheaper than Opus right now.
You cannot just look at the price tags for these models, you must eval and see the price per task. In our previous eval rounds Sol was more expensive than Opus (with its original price), took much longer, and provided worse results. Kimi does not have these issues, it's just as good as Opus with a smaller price tag.
China's 50 Cent Party being a real and noticeable thing (and the two biggest things they like to shill is open weight Chinese models and the futility of resisting a Taiwan invasion), I have to take things like this with a healthy dose of skepticism without corroborating data, since independent evals didn't show the price per task lead you're showing.
If there's independent data showing this feel free to share a link, I haven't seen it. DeepSWE has been most closely matching what I see in my own use.
> And these models are not going away, nor their prices going up [...]
Well, DeepSeek just raised prices.
And Fireworks did not yet. They are still under the limit of not feasible to self host... Let's see if other providers follow DeepSeek with their flash pricing.
After using Claude for a long time, I tested Sol 5.6 for the first time today. Love it, its an incredibly capable model and uses far fewer tokens/time thinking. Its what I imagine Fable would be if I haven't been downgraded on every conversation - even after completing the verification program. I think I may cancel my Claude subscription finally.
I switched from claude to gpt when 5.6 came out for the same reasons. I don't understand why so many people are still using claude when GPT and open source are so much better.
I think Fable's dominance is overstated. It definitely has the lead, but quantifying what that lead actually is is really hard. I'm using GPT 5.6 Sol to do some shit that I personally would consider "crazy" - low level undocumented hardware driver alchemy, reverse engineering highly obfuscated code, even a bit of screwing around with a rendering engine in Vulkan, really just about the most complex tasks I can get any model to do, and it does great. For the more advanced stuff, it definitely needs the effort bumped. But even with the effort bumped, the token usage really doesn't seem to skyrocket too badly until at least you hit xhigh and max, which really only seem to be necessary if you are doing genuine crazy stuff, so it's not that bad. I did similar stuff with Fable. In fact, I went directly from an Anthropic subscription with Fable to an OpenAI subscription with Sol, more or less, and it really felt pretty seamless. If anything, I was thrilled to realize how much I actually preferred Codex CLI, to the point where I started using it at work too.
Fable seems to be generally more impressive at outputting one-shot web apps. I'm not really saying that to try to downplay what Fable can do, it's just that if I compare the two, this is one of the few definitely noticeable areas that you can easily demonstrate. Obviously, one-shotting programs is much better as a demonstration of a model's capabilities than it is practically useful (not that it is useless, but hopefully my point is understood).
However, whatever Fable truly is better at, one thing I really like about GPT 5.6 Sol is even harder to quantify: taste. GPT 5.6 Sol outputs are still LLM outputs and they contain many things that people would probably consider "Claude-isms" for better or worse, but overall I really prefer the GPT 5.6 Sol output. I find it to be generally more tasteful. Hard to quantify, but when talking to people I've had enough people seemingly agree with me to convince me that it really is true.
I used Sol to extract the remaining decryption keys from the Super Mario Maker 2 (Switch) game files. Someone had previously extracted all the keys from the original release, but not any of the new ones from updates. Not only did it succeed, but it helped me understand the data sufficiently to add support for “Super World” rendering to my level viewer (which I made back in 2021), eg the little widget at the top of https://www.smm2-viewer.com/players/B16-306-GVG
I was very pleasantly surprised to find Sol wasn’t obstructive over what was clearly a very grey area endeavour.
Interesting. I've been wishing that old 'Stars!' game from the 90s would play easily on modern systems. I'd love it if we'd got to the point where I could point Codex at a folder with an ISO from my CD of the game and tell it to go reverse engineer it all for understanding of game mechanics, then go recreate it in a modern language capable of running cross-platform. Scarcely any need to improve on graphics, it could even be a PWA.
There's people that have tried to contact Jeff McBride and follow the IP trail but the IP is currently owned by a company that went defunct. Not sold, but no one is even bothering to register its LLC any more, it's simply dead.
I was having it look at creating a driver for some old scanner and it actively looked up exactly where that gray area for my country was wrt decompilation.
Fable is almost unusable for anything but super boring mainstream stuff. I was getting safeguard flagged so often I’ve significantly reduced my usage out of fear they will blacklist/ban me.
Some of the topics it’s flagged have been hard for me to understand what it seeing that can be remotely concerning in my requests.
I cancelled my Claude max subscription. Somehow every query I sent was flagged as bio or chem, even pure mathematics questions. Not going to waste money paying for a “max” subscription that won’t ever let me use the top tier model…
Sol is great and has never blocked a request, and generally gives great answers. Happily switched over to it now.
The safety is really funny to me. I ask it a lot of extreme stuff and it goes through, but I ask it mundane stuff and hit the filters all the time.
Reminds me of the URL blocking of my company. Nature.com is being blocked but I can access a ton of super sketchy download sites.
It has learned a little too well how it works in human society.
The dangerous part isn't that a model refuses extreme requests. It's when mundane requests become unpredictable enough that you stop trusting the model.
I've gotten flagged for asking questions about tokens and tensors. That makes me believe it's not about safety, it's about protecting their turf. I cancelled my subscription - same fear about getting flagged too much leading to a ban.
They said they also block usage of Claude models to build ML models.
Which is definitely protecting their turf, but also probably a little bit hiding their “RSI” abilities for competitive reasons. My theory is that a lot of “safety blocking” is actually WIP training of new business directions. Anthropic has started hiring biologists and has opened a preview of a “Claude code for bioinformatics”. I’m guessing they’re tweaking their bioinformatics market play, and block “bio safety” requests so competitors can’t learn about their training.
OpenAI and Kimi are both pretty okay alternatives! I guess GLM 5.3 on Max reasoning as well but for more limited domains.
Have they made Sol do less unwanted autonomy than the previous Codex models did?
I feel like those examples are considered difficult because they're niche topics, but aren't actually all that difficult in a general sense. What I consider truly difficult are things like taking a ticket and implementing it in a preexisting codebase, using a clean and reasonable design that fits the existing style and makes sense to a human, and avoids the footguns I learned by working with the codebase for over a day.
If you said this in 2025 I would've 100% understood, but to be honest getting AI models to do a pretty good job on day-to-day ticket work has become so boring that we don't even bother using the top tier models and higher effort slots for that anymore. I personally wind up tweaking the results a lot and recursively having fresh agents review the diff, but that's just because I'm picky; in a lot of cases the first diff is actually pretty damn decent.
Compared to what I am doing at home experimentally, I feel like day-to-day work is absolutely nothing. Not only am I also working with existing codebases in my experimental prototyping, but I am also doing things vastly more complex with vastly harder constraints.
All non-trivial code terra has generated for me has had at least one serious bug in it. Typically caught by a review from myself or Sol.
But I wouldn't trust lower tier models for end to end solutions.
This is true in some sense.
Getting the AI to output code that you like is difficult.
As an example, let's say in React you have a "useLocale()" hook.
The AI will happily pass down locale as a prop to 5 child components instead of just calling the hook in the component.
A review from another model did not flag such stylistic issues either.
I believe that the latest models are very good at functionally achieving the goal, but still have poor taste for UX or code quality.
The most productive use of AI for software development happens in an environment where you do not review the code but test the UX end to end.
I use the AGENTS.md to show it how i want the code to look like. Something like "when implementing hooks adhere to the guidelines in docs/react-hooks.md". And then react-hooks describes your heuristics and what you consider best practices. There is a clear difference in code quality for me when using codex with a well crafted AGENTS.md vs. without one, you can run the experiment yourself pretty easily. As I mentioned in another comment, I think Claude poisoned users to stop relying on their Claude.md files and new codex users might be surprised at how well it adheres to guidelines.
We use skills for similar purposes, with positive and negative examples.
I think it sometimes worked, for example for testing preferences, but sometimes it did not.
Could be a problem with the harness also.
In any case, I feel that it's a bit playing whac-a-mole with explicit rules for things that a more intelligent model should do by default.
FYI I run it consistently in xhigh regardless of difficulty of the task at hand. I remember high being very fast, but I'd rather wait a bit more and get better output. AIs are insanely fast compared to me anyway, even on xhigh. Consumes more usage, but even at 100 EUR/m I don't hit limits.
After hitting the session limit on my company's plan so many times with Claude when I was using it, I mostly keep Codex on "high" rather than "xhigh" as a way to leave the tokens for my more ambitious coworkers. It's possible that having it higher might end up with better output, but so far at least I've yet to see a way to get any model to do 100% of what I need up front without any need for me to make changes that end up being more tedious to do via interaction than by hand, and it doesn't feel worth spending a bunch more tokens trying to figure out how to better communicate to it up front how the dominoes get set up so they fall in place properly the next time.
To be fair, I actually do run xhigh as my default. However, for the first time in my experience of trying and using LLMs, with Sol.. sometimes I feel confident enough to set the effort level to "Low". I just had Sol prototype some AWS stuff on low earlier. Great result, did exactly what I wanted.
How do you handle context limits? With more thinking tokens you fill it up earlier. Compaction degrades performance too. What's your strategy?
Initially I was planning heavily around context limits, but I've learned to just ignore it completely. Compaction is seamless for me. If details are lost in compaction, the model just re-reads what's needed. My conclusion is that at least for Sol, the summaries (which I've never seen) must be amazing. Every now and then a detail gets lost and I have to repeat it. I don't think there is performance degration, because the model is smart enough to re-read relevant files as needed.
And Mai-Code-1.1-Flash seems like a really good cooperative player to GPT 5.6 Sol. You get Sol to help you make a detailed plan, and Mai codes it up and you can get pretty decent code out the other end without too many tokens if you are careful.
How detailed of a plan? Are you including code snippets or just behavior and letting the lesser model decide how to implement?
Why wouldn’t you use Luna for that? It’s super cheap.
> Why wouldn’t you use Luna for that? It’s super cheap.
MAI also offers a ultra cheap version that's competitive with Luna.
So much so that the models look like they were designed by a product manager explicitly to eat away OpenAI's market share.
Vscode even pushed them quite hard onto users with the latest release, going to the extent of putting up a modal to convince users to try them out.
Taste I suppose?
AI-pilled obsession with "taste" is bordering on insanity
It's just vibes
It's vibes all the way down.
It's... really just vibes?
Always has been.
things that are alchemical are rarely alchemy. That is to say things are very fiddly but stick a room of monkeys on typewriters, a schizophrenic developer with HolyC and adderall or an LLM, persistence is the key to many of these things like drivers, extracting keys from vintage security domains, etc. Dropping into xdd to a human is a chore, not for an LLM.
Although I am not exactly sure what you mean, I am not really claiming it is doing anything I couldn't do - but yes, it does so with much less effort. For example, I can have it set up probes and tracing on Linux that I personally would have to consult documentation to do. It might not even have to consult the documentation due to having the information on-tap, but even if it does, it's nothing that would cause it any fatigue, it's just going to keep moving forward in a loop until it is satisfied that it meets the criteria. I could've done all of this alone - I really could have. I just would not have. Being able to do something 10 times faster or with 10 times less effort is, in some senses, sometimes more impactful than being able to do entirely new things you couldn't do before.
This is a really good point that I definitely failed to grasp when first hearing about these tools. At least for me, the best way to use these tools is as a way to free myself from having to spend time thinking about the things that aren't worthwhile so I can focus on the things that truly are. I've had times in my life spending hours reading documentation and googling random things to try to tease out the correct sequence of commands or the exact right shape of an API to be able to make things work to know that it doesn't make me more productive to do that myself rather than point an LLM at the thing and let it spit out the answer after a few minutes. Meanwhile, I can spend that time thinking about what comes next, or what the correct way to take that one-off output and abstract it to something that can be used meaningfully in more flexible ways.
The only obvious objection I can think of to this line of thinking (at least from a technical perspective) is "how does someone build up the knowledge to be able to use a tool effectively in that way if not by doing things by hand at first?" The honest answer that is "I don't know, but that's also pretty much exactly the type of thing my employers have never been paying me to solve in the first place". Even just a decade into my career, there have already been plenty of times in my career I've struggle to convince people that we should do stuff in a way that won't bite us in the ass a month or two down the line, and in the times I've managed to succeed, it's usually only by putting in more of my own time and effort to make the initial investment seem more palatable. Luckily right now I'm not in one of those times when I'm having to go full throttle to keep the lights on a few months from now, but I don't have enough fuel in reserves to work on a plan for when we need to build a new rocket in another ten years. Maybe ask me next month.
Sol is way too eager to hone in on small details and ends up with massive over-engineering. Fable does it too - to be fair - but noticeably less.
After extensively using both on Max 20x plans, I've concluded that Fable is better for problem solving and coding, whereas Sol 5.6 Ultra shines in debugging specific issues: tackle a problem with Fable then leverage Sol to clean up, double check, or fix specific issues.
Fable (imo) had the edge on the $200 plan, but after this 50% reduction I'd say Codex is better value by far and there's no contest.
---
Using Fable as the orchestrator and delegating tasks to Sol 5.6 Ultra via the codex plugin in Claude Code yielded good results, but still there was a lot more over-engineering (thus time and tokens spent) than Fable by itself would've done.
Both models suffer from doing-too-much. But both models are fundamentally really smart and knowledgeable. I think it's really close and pricing cuts really spice things up for us consumers! Sol is a clear winner in the value department and the $100 plan is enticing!
---
*Claude Code usage is reducing by 33% in 2 days, Wednesday August 19... cmon anthropic: clau.de/cc-50-promo
Yeah, I very much agree on this. I think Sol and Fable code quality is on par. Maybe Fable is just a tiny bit better, but Sol compensates with its ability to work through things, while Fable, in my experience, generally tends to avoid solving problems that require many LOC.
However, I think these are very different models in terms of orchestration. Long-horizon tasks are way more predictable with Fable. It just doesn't lose track of details. Thus I ended up building a small wrapper around Pi (where I run Sol) so that CC can delegate via background tasks, automatically wait for completion, and do what was one of the most effective parts - steer Sol toward simplicity, getting Sol out of code-review infinite loops (Pi calls for Codex review to ship better, but generally gets stuck on P2 and results in vastly overengineered work).
One of the worst experiments was enforcing coverage at 100%. Only Sol, with an enormous amount of code and significant pushback (on architecture decisions) to Fable, was able to reach it. It made me think this is somehow related to overengineering in general, so that instructions on acceptance criteria in claude.md plus proper DX (e.g., Lefthook) actually led to okay results. It mostly helped that responsibilities were clearly split: Fable designs architecture, Sol handles coding and debugging.
Sol w/ Effort -> Low
It's great, don't get me wrong, but so is Fable. I'm just comparing the long-horizon task performance between the two at the same or similar effort levels.
Given the 50% discount on Sol and how smart it is, yeah it's unprecedented value. If you only want to use low effort, there's a clear winner here on value and it's not even close!
Interesting the use of Max and Ultra. I don’t doubt the complexity, but would someone use Max or Ultra on Typescript or Go, for example?
Is it more about just avoiding any mistakes? Seems like that would be costly when medium or high would work fine?
*The "Max" I referred to was the plan tier, not the effort level btw
For small tasks, you can just use something like low or medium effort and it can usually avoid mistakes; after all, the model will test the code anyways and can do some baseline level of iterating.
In regards to cost, we need to acknowledge how generous OpenAI was in the last couple months with Codex usage credits (no weekly limits) and usage resets. It afforded me many a dive with Codex! Yes it uses more tokens, but sometimes it's worth it -- just depends on what you're working on.
Finally, Ultra(code) isn't that bad when it comes to cached tokens. I think folks overstate the general token usage of ultra effort on both providers.
---
Both models are great at green-fielding a project when given detailed specs.
Both models overthink too liberally (imo) during these larger multi-shots. Sol overthinks more than Fable.
Both models are really smart and perform great for general knowledge and regular coding tasks.
5.6 Sol is a joy to use for "daily chat" as well. Compared to earlier OpenAI models it catches and corrects its mistakes very reliably. It also seems way smarter in tuning its replies to areas I am more/less knowledgeable about (i.e. when I ask it a law question, it assumes I know as much as a toddler which is true, but on political topics it more easily throws around terminology) and including analogies. On medium thinking, it's a very good compromise between speed and quality.
Fable feels less cumbersome to work with, but it is SO DAMN ANNOYING with the refusals that I'm leaning more and more on Sol, and very much looking forward to GPT6. Just seems like Anthropic is trying their hardest to ruin their reputation and user experience.
Remember, Dario knows best
I’ve never had it refuse anything. Even vulnerability searching in my codebase.
Yeah I feel like I'm living in a different dimension than these people. I wonder what they're working on. I've literally never had it refuse everything and I max out my 20x plan every week
Can I ask which country you're in? I have a theory that the safeguards differ depending on the user's country.
I'm in Australia, and Fable downgrades to Opus when testing for bugs in memory in a legacy C code base. If Fable starts taking initiative and writes a test case that involves writing to a null pointer, that's the end of the conversation.
As one random example, today I had it hit a refusal loop when adding a country selector dropdown to a form, presumably because it contained a “bad” country name? I hit refusals at least 2-3 times per day, sometimes many more. The worst part is it is often right in the middle of a multi-stage task, so the only option is really to switch to Opus 5 and let it defecate its absurdly verbose comments all over the rest of the edits in the turn and hope it doesn’t go on one of its tangents, then have Sol do damage control. Oh and I got approved for their “cyber verification program” blessing, which comically does absolutely nothing for Fable.
I too switched to OpenAI after I got sick of Anthropic's constant "safety" downgrades. Sol is definitely a breath of fresh air.
> even after completing the verification program
Was it easy to complete it?
I ended up in some weird state where I can't even attempt the verification at all. Opened the Persona tab once, closed it and then it never opened ever again. It says a verification precheck failed.
Even without TAC, Sol doesn't seem to get blocked very often. Fable would downgrade to Opus if I looked at it wrong.
Fable is still the best there is. Sol close second but I find it gets way to stuck on details.
Also Opus 5 is fine if your codebase is simple.
And you arent writing English
I was pleasantly surprised to find that the GPT models are much stricter in adhering to my AGENTS.md guidelines and heuristics than Claude.
I'm also in Anthropic Cyber Verification Program, but they specifically exclude Fable, just goes up to Opus 5.
I hear you on the downgrades, I'm 13/13 on downgrades, and last downgraded me to Sonnet for asking for reasoning chain.
I cancelled my subscription recently and moved to Sol. So far - it has been a great experience. The only aspect where Fable/Claude is better I feel is doing some research from the web and summarising the facts.
I generally prefer Fable but in my experience Sol is a much better web researcher
Sol is my daily driver but there are still times I reach for Fable when Sol doesn’t cut it. Just yesterday for example, I was trying to build a self-modifying hot-reloaded agent harness in Elixir for fun and Sol just kept doing silly things like thin wrappers and unnecessary abstractions. Fable handled the task elegantly. Sol is really good as a reviewer for finding bugs due to its thoroughness however.
It's the complete opposite for me. The model might be the worst model I have ever used when compared to other models in the class. You just can't get it not to just write the most enterprise over complex over engineered solutions for every little thing you ask it to do.
It the first model to actually make me pissed off to use AI. I absolutely hate the model so much.
I don't even want to see the codebases this model is fucking up.
It might just be good at finding bugs that about it. That all I would ever use it for just because it works harder than Claude models.
output token efficiency bruv. u never go wrong with it
Used claude since 7/2025. Switched to codex after fable got blocked. It was still 5.5 but I knew they had to come up with something. As soon as I switched, wow. It wasn't super intelligent, but it was stable. Every day it was the same performance. This consistency is definitely worth paying for.
you should check out the codex desktop app. people who've been using claude code for a long time will surely be surprised.
Yeah at this point claude is overrated, overly expensive, weird writing style (elliptical), and the worst part is the aggressive guardrails that even normal convos get interrupted, meanwhile openAI is still I would say at the normal balance, if you ask something too obvious or direct it will stop you other than that, it work flawlessly, plus, I have yet to hit the limit despite heavily using it these past weeks.
> if you ask something too obvious or direct it will stop you other than that, it work flawlessly
What on earth are you asking it?
sol is much better imho than Fable but i can understand if they will perform wildly different for different people with different levels of expertise aswell as different needs. I dislike fable myself it doesnt really work for me.
Sol also doesnt _really_ work but it sort of tricks me into thinking it does more convincingly :p.
cancelled my subscriptions few days ago. (was on 100$ ones, not sure if there is diff in quality for higher tiers or not.. there might be that too).
what i hate the most is that they will make any obvious mistake you do not tell them to avoid. then on the next plan to fix it, your token limit is hit at step 4/5 -_-. Both models seem incredibly good at that mostly...
for tasks outside of coding and program design i do find them quite useful. like devops crap. maybe because i hate that, i like their help there more.
The final straw for Claude was its refusal to give me a list of the most recent rapes reported by the BBC and basic information about them (location, date, names, just things reported in mainstream media). It outright REFUSED to complete this task.
I will not be told what I can and can't do by AI and I will no longer be supporting American companies run by despicable people. GPT only gets my money right now because its so fast and cheap but I'll be back to Chinese models in no time.
Claude as a harness at all really spends too much time before giving user feedback
Its a crutch that is no longer competitive
I have a "strategy / life-coach" project, and was surprised at how much better Sol is than Fable on it, as I've found Fable to have the edge for most things for me so far. But Sol: questions were better, insight was better, it got the brief better.
Sol has held stuff for a while to do the same sort of hazard checks I assume Fable is doing, but it always releases them. I think that's the better way to handle it rather than preventing me from seeing how far I can get generating schematics to use in Minecraft. Currently: a mostly normal voxel house.
I recently tried Claude again after several months, to see if it was any better at something Codex has been struggling with…
They STILL don't have an option to "Sign in with Apple" on the website, but they do for Google??!? (and on iPhone of course)
Screw that asinine UX
(and no it wasn't better than Codex at this particular task)
That was an issue at least a year ago. I had signed up for a claude account on my iPhone and then wanted to sign in on my laptop but nope, not possible. Insane they still haven't fixed it.
Can somebody at Anthropic tag claude in slack or whatever goofy shit you do and ask it to add Apple OAuth to your website? Clearly humans aren't testing it.
I signed up on iOS, Sign In with Apple, cause I don't go around giving random companies my actual email if I can help it
and sure enough, I was right to do so: They don't even let you remove your payment method afterwards. Every other store, Steam etc., lets you.
No way I have enough trust to install their desktop app after that, so I just want to try it through their website..
Can Sign In with Google, but not with Apple
so you gotta open the Passwords app, copy your random email, paste into the website, then copy the OTP from your email..
It's been that way for at least a year
and the AI itself hasn't been that hot compared to ChatGPT/Codex either: {pic incoming}
All the Claude hype posted on HN seems like a case of the emperor with no clothes
There is no Claude hype on HN, HN has more vitriol for Claude than any online forum I've seen
This sure looks like a race to the bottom to me, and I love it.
If Sol isn't the best model, it is up there...
You don't cut the price of the best model for no reason...
> This sure looks like a race to the bottom
Always has been. My prediction is that both OpenAI and Claude will go bust unless they deliver a killer product. And unlike scrappy startups, they have a pretty serious deadline because creditors will come a-knockin'.
There's little to no functional difference between Kimi, Qwen, Sol, Opus, etc. All flagship models are within like 1-5% of each other and the real moat will be what's always been the hard part: making a good product.
> All flagship models are within like 1-5% of each other
Don't know about that.
I'm using code review of my lone lisp project as a benchmark. It's a massive parallel code review where a coordinator cuts up the codebase into sections and dispatches agents to consider each part from different perspectives like quality, maintainability, consistency, correctness, rigor, etc.
Ran a complete Fable/max code review. Took over a month on a subscription. Now I've switched to OpenAI and am repeating the exact same review with Sol/max.
It's still not done yet but preliminary findings suggest Sol can only reproduce 70-90% of Fable's findings. So I think these models aren't as close as we've been led to believe.
Check the remainder for hallucination for sure.
The problem is that most of the volume doesn't come from proprietary products, it comes from API use which has no stickiness.
Claude already has a killer product (claude.ai/chat is a Swiss army knife) but just relying on people typing stuff into chat is not enough to sustain the company.
The other strategy is entrenching yourself as the LLM of choice into existing products (like ChatGPT is on Apple products).
There is a massive difference even between Opus and Fable, same provider, before various harnesses and other optimizations come into play. Don't be deceived by rankings and benchmarks, try for yourself.
> All flagship models are within like 1-5% of each other
Depends on your use case. the Chinese models are not there yet.
If you believe https://artificialanalysis.ai/
This is basically undercutting KimiK3 and Grok 4.6 where previously utilised gad soke advantages but was a step more expensive
You do it if you can afford to do it and your competitor can't.
Who are OpenRouter’s competitors?
OpenRouter doesn't decide on the pricing.
They do decide on the 5% markup. But as far as I can tell, all other routers just match 5%. Not sure what they're competing on.
Yes but Sol dropping in price by 50% is not OpenRouter deciding. It's OpenAI.
I don’t think that’s true. OpenAI docs don’t have this price change. I assume they’d be the source for this post if it was true. The banner on OpenRouter for me says Gemini 3.7 discounted for a limited time, but if I click through that I get to this page: https://openrouter.ai/models?discount=true
That shows a bunch of models, including Sol, with a discount. None of them say how long it’s for, but I’d assume in all their cases it’s for a limited time as the banner said, and only on OpenRouter.
OpenRouter margins are not 50%.
As I understand OAI is offering discount only for users using the model via OR. Not sure why. Maybe they want OR users to try the model and switch to OAI subscription or something.
It might also be worth noting that the discount is only if OpenAI itself is used as the provider via OpenRouter. The discount does not exist for Azure or Amazon hosted Sol.
I posted elsewhere, but the Azure uptime & performance for Sol is truly dire. OpenAI is offering 5x faster latency, 4x faster tokens generation, and vastly better uptime (Azure US has only 87% uptime), all for 50% of the price now. I assume the pricing is to compete with other shiny new models (Grok, Qwen etc), but it might also be to cut-off a truly poorly performing Microsoft hosting experience.
Yes, but this discount must be coming from OpenAI and not Open Router.
That kind of is a reason.
The lower the merrier.
Competition is good for users.
Sorry I meant to say lower, price that is, corrected.
Where is the official source for this?
OpenAI's docs still show non-discounted pricing https://developers.openai.com/api/docs/models/gpt-5.6-sol
It's discounted if used via OpenRouter, not the official API.
Doesn't seem like a smart business move.
You're literally encouraging someone else to come in and steal your customer base,
A surprising amount of companies sell exact the same product through different channels for different prices. A good example is Apple.
Multiple times a year, retailers here in Australia have co-ordinated sales on Apple products. Apple.com or their retail stores don't have these sales.
But they're clearly Apple-funded when competing retailers launch the same sales on the same days; and the margins aren't enough for retailers to take a loss.
Or it's "normal" market segmentation. OpenRouter users are more price sensitive in general, also a lot of enterprise users who can't switch easily are using the official API (or Bedrock or Azure) and you want to squeeze them as much as you can.
Literally all OEMs do this, if you go to Apple site you won't see any sales, but Amazon and other retailers will have stuff at 20% off regularly
This imbalance of exchange means that OpenAI is getting something from this deal. Question what is exactly.
Seems reasonably clear to me? Potentially bringing in more customers who use OpenRouter for trying out all the models with rapid switching, enticing them to use Sol. And anyone being routed on price will immediately be switched to OpenAI's servers instead.
It also seems to be providing a vastly better user experience - Azure has less than 99% uptime (Azure USA only has 87% uptime), latency of 20 - 30 seconds, and a mere 8 tokens per second. OpenAI is offering 32 tokens per second (4x faster), 4 seconds latency (5x faster), and all for half the price of what Microsoft is charging for a vastly inferior experience.
Data taken from this page:
https://openrouter.ai/openai/gpt-5.6-sol#providers
So the title is misleading, intentionally.
That's weird.
Maybe they (oai) want to pump their marketshare on openrouter lol
Why is that weird?
How is that not weird?
Has OpenAI struck a deal with openrouter and that's why we're seeing preferred pricing?
Is openrouter taking a loss on sol API calls to grow adoption?
How temporary is the reduction in price?
Selling cheaper through what should be a minor third party, than through the first, party is super weird in commerce.
Luna saw a huge jump after the price cut and is one of the more competitive models at the new price on openrouter.
Maybe they want to see how much market they can grab with Sol?
This might help but there are already cheaper models with Sol's intelligence more or less, the most notable being Grok 4.6 at $6/m which makes it a tougher sell
It's really only between Anthropic and OpenAI for many of my use cases, since I have a Zero Data Retention agreement with both. I'm not trusting random inference providers and especially not Elmo with sensitive data.
or Tinfoil [0]? They serve open models with container integrity attested by Nvidia/AMD enclaves. Every cloud provider offers this of course, but not usually in a way that can be shared between distrusting users for economical inference. It still relies on the open-source containers being secure, and there's probably hardware sidechannels and stuff, but personally (ie privacy not liability) I trust it more than a contract
[0] https://tinfoil.sh
I loaded credits into Tinfoil the other day and have had zero issues, thanks to smart folks around here and the security space recommending it.
It's now my open weight inference provider of choice, since on top of the privacy/security characteristics it's also reasonably cheap.
plenty of companies offer ZDR and are just as random as OpenAI and Anthropic in their age
Would love a few names. Many of them fail to provide good uptime for large scale jobs or don’t have batch apis at all in my research.
DigitalOcean does inference, has good uptime and offers ZDR. They are not an AI first/inference first company. They have been serving cloud products for almost 2 decades.
Batch API is a bit harder, not many models/providers support Batch. It primarily is only the Gemini/ChatGPT/Claude models that do. DigitalOcean does support a 50% discount on Batch API via them directly, not listed on OpenRouter.
[1] https://docs.digitalocean.com/products/inference/how-to/use-...
You do not have a ZDR with Anthropic.
For Mythos and even Fable they require prompt retention on their end.
edit: or more precisely if you want to access Mythos/Fable ZDR does not apply, and depending on config the exclusion can affect other models.
Since when does Grok 4.6 have Sol 5.6's intelligence? I don't believe it.
I do have free sol and cursor ultra for 200 I prefer grok over sol, they are equally capable but grok is faster
I use Grok 4.6 every day; it's good, but it's not Opus 5 or Sol 5.6. The gap is closing, though.
Why not?
Because it’s good on benchmarks but not on real usage?
But OP said they've never used it. How would they know?
It doesn't.
I wonder if xAI is A/B testing routing some difficult grok 4.6 queries to Sol to seed some true believers.
I have switched to Chagpt sub now after only using Claude for coding. You get more value for your money and feels like codex has reached Claude code performance in coding (the reason for using Claude) regular plus account allows you to have access to their most powerful model, image generation and asking questions is better because you can use sol but in instant mode and it feels smarter and faster. And finally codex usage limits are better than the Claude daily 5h limit. And codex feels faster although Claude code had more features
> codex has reached Claude code performance in coding
Codex has always beaten claude in coding benchmarks, hasn't it?
I switched as I felt Codex was on a par with Opus, but the chat responses from Sol are just more intelligible than the word soup I've been getting from Opus. I wonder if Opus could be prompted to respond in simpler prose via agents.md
Ah, so I am not the only one struggling with deciphering Opus writing style. At times I feel dumb as a rock because I read the same passage like 5 times and I still don't get it.
> I wonder if Opus could be prompted to respond in simpler prose via agents.md
God knows I've tried. I've got a variant of the ASD-STE100 trick which does the job, mostly, at the start … but get to about 100k of context and it goes out the window.
The model's personality is too strong for simple suggestion, alas.
I have also noticed that I am increasingly struggling to read what LLMs are writing, finding it incomprehensible half the time. I stole Matt Pocock's line of "When reporting information to me, be extremely concise and sacrifice grammar for the sake of concision." for the agents.md It makes it a little better. I also specify to use https://github.com/AminBlg/SimpleEnglish/ for all writing it does including code comments. It all feels like a bandaids but that seems to be the best we can do right now.
I used over a billion tokens per day of gpt-5.6 sol xhigh starting last Wednesday through Sunday before reaching my reset limit. The $200 pro plan is still the best deal.
I spent $800 in a few hours when my sub maxed out because I was trying to get something done and had a long car ride to let it churn.
Their api pricing is absurdly expensive.
> Their api pricing is absurdly expensive.
I assume at this point that it subsidizes subscriptions.
Subscriptions are a mechanism to attract developers, who then advocate that their company should use the API.
yes, it absolutely does.
I've gotten more work done on a second chatgpt pro $100/mo subscription than I did with ~$150 of paying for usage through the app.
Massively subsidized. As soon as my Claude switches from subscription to overage I have to tap out quickly.
This is a big part of the reason I went local-only. Subscription limits are horrible for having a decent workflow.
Because API pricing is for corporations and subscriptions are for consumers.
Are we supposed to just shrug at the idea of businesses paying 10x more for raw materials than consumers? How long can this go on?
Hopefully for a long time this is like the beginnings of a vac startup, enjoy while you can before the enshitification comes
I thought corporations were meant to be smart and get b2b / volume discounts - not pay 5-10x what the man in the street is paying.
FOMO. This won't last forever.
A billion a day? How many agents are you running?
with ultracode, it goes fast. I can easily get to a billion on a busy day.
That's wild. I was gonna say 3-5 billion a month is more reasonable summed across all token types.
I am also on the pro $200 plan, the limit is high enough to do what I want for a week, and binge run Ultra Fast last day to use the remaining credits.
I have mine churning like butter and I'm rarely hitting a billion tokens per day, what's your workflow look like?
Pretty basic. The codex app with one conversation per project and several running simultaneously all hours. I’m going for max caching that way and it never gets lost even with compaction somehow. Each has a plan with milestones to keep up to date and a thin agents file. I check in on them in the Remote app. Use case is protocol and control reverse engineering of audio hardware. I think they must be identifying the heavy use agent sessions and cranking up their cache lives so it’s not a big deal for them.
I'd like to see a benchmark on this specific topic: Reverse engineer the hardware protocol from a driver, or just migrate a driver from one OS to another.
A billion tokens a day is 11,000 tokens a second sustained. How many tokens per second are you getting off of GPT 5.6 Sol per project?
Ultra mode spins up many sub-agents. On a particularly challenging task, I’ve had as many as 29 agents working at one time.
Also if you don’t specify, most end up being the same as the parent model which is pretty wasteful.
I engineered a skill that spins up Terra High agents for most sub-agents, resorting to Sol Medium for technical research and Luna High for code/in-project research tasks.
On a slightly different topic, Luna Max is incredibly capable and doesn’t use as much quota (Luna tokens are dirt cheap).
Most importantly, are you seeing a return on investment for time and ultimate outcome?
No one can judge the enjoyment, learning, and hobby aspects. Just wondering if there is an end goal for that much overall expenditure (time, money, energy, etc.)
I absolutely see a return. I do not pay for this (we have a corporate gateway), for the price of a junior developer I can get 3-4-5 senior developer's work done. it's insane value.
Often people are counting all tokens, including cached input tokens, for those more impressive "billions of tokens" quotes.
Ah, thanks, I'd missed that nuance in the other reply!
Some people just do crazy stuff. For example this now ex yc guy who said he has agents constantly scanning Sf govt apis and forming dashboards just because
Is anyone hitting caps without agents or API usage? Seems very difficult.
Yeah I am, building large software with a vision - requirements - architecture - plan - code workflow. One Claude max account is enough to work on one, maybe two of those at a time (call it 15B tokens/month per project)
I’m exclusively using ultra and I run out in 3-4 days consistently. Those resets are great but I’ve noticed they like to cluster them at the start of the cycle, would be better if they spaced them out more.
A billion tokens per day?? Plausible estimates put the energy use at about 0.001 Wh/token, which means you're using 1000 kWh/day in electricity, just to generate slop. That's about the same as 50-100 houses. 300kg of CO2 per day - roughly the same as flying from London to New York every three days.
I think on average AI energy usage is not as big a deal as everyone is panicking about, but your usage is truly absurd and I don't know how you can live with that. It's immoral.
That’s for i/o tokens, mostly output. 90-98% is cache read usually, so you can divide electricity use by 10 at least.
As for co2, it depends on the provider, it could be way lower as well.
As for ethics, you don’t know what he works on, and how effectively - he might be saving 10x that much of co2 for the planet.
I'm glad someone is voicing this. Overconsumption at that level is not defensible. However if they meant cached tokens so it's not that bad.
I can't speak for that guy, but I'm a physicist and work in clean energy... So it's not too hard! That said, I usually am closer to 10M on days I do heavy coding, so not nearly that bad.
Token caching is a thing
You really think OpenAI is selling $1500/mo of electricity (at $0.05/kwh) for $200/mo?
I’m guessing that Wh/token estimate is several orders of magnitude too high.
They certainly could be using that much electricity at a loss based on their profitability, which doesn’t exist.
Leaked financial documents from 2025 show the company reported an operating loss of approximately $20.9 billion against $13.1 billion in revenue.
As a company... but that includes things like research costs, model training etc. to determine if they're selling electricity at a loss you should look at inference costs bc that's the "thing" they're selling
1. You do not know what they're using it for. 2. Get off your high horse please. 3. Immoral my ass.
> Immoral my ass.
Is any amount of tokenmaxxing moral?
Is any form of energy usage moral?
Most tokens are cached.
"slop"? come on. we're not in 2020 anymore, Dorothy.
I can’t sign up for that. I tried authorizing Codex a couple days ago. For some reason, their system says my phone number has been used for verification 3 times even though it definitely has not. I’ve had this phone number for over 20 years. OpenAI support is useless. They just keep repeating the policy without actually helping me.
Use TextVerified, load up like $5 of credit and OAI verification is like $1.00. Then when your account is made, ensure 2FA/passkey is setup then you don't need to worry about the phone number.
Thanks for that tip!
Yes I filed a support ticket with them and explained that their system is broken and they just did not care. I explained how it was impossible for me to use it 3 times already as I've only made 2 chatgpt accounts EVER, and only recalling entering my phone number for one of the two chatgpt accounts. I told them that this issue locked me out of codex and chatgpt for work and they weren't willing to do anything about it. Totally useless support.
I ended up borrowing my gf's phone number just so I could get access for work. Ridiculous
Get a burner and use it? If you're spending $200/mo on something, $40 or whatever for a burner phone seems like a pretty cheap price.
Oh no fuck that, business 101 is make sure that your checkout page works. There is plenty of competition in this sector, take your money elsewhere.
You can get a phone number online for a few dollars.
Historically those are less useful because some of the verification systems require a real phone number and that your name is associated with the account, depending on what and how they verify. It's annoying, I use a google voice number as my primary, and it often gets rejected.
Good2Go is $5/mo for a real SIM with unlimited talk/text + 1GB data.
If they can cut the price of Sol by 50% and the price of Luna by 80%, then the original price might have carried a massive operating margin. They might still be serving the models at a profit after these price cuts, but we will never know.
I don’t think there’s a real answer for this. Margin depends on whatever number the accounting department wants to make up.
Do you include research and training costs? Of all models or only the ones being served? What percent of the R&D budget do you allocate to inference? What about data center capacity? Do you count future commitments? All the circular financing deals? Do you count employee equity grants as costs? At what valuation?
We have a simple definition for this: COGS.
We also have another solution for "whatever accounting decides": generally accepted accounting practices. It's far from perfect, but GAAP figures are what you should be looking at; not "adjusted GAAP" or whatever invention.
Or they have gotten new asics and can do now inference way cheaper
Why would they infer faster with new shoes?
It is a well-accepted fact new shoes make you faster. Current science suggests it is due to the lighter weight from lack of dirt, though there is a competing theory which says it's an optical illusion due to the fact the pure white streaks resemble speedforce.
OpenAI didn't cut the price of Sol by 50% like they did with Luna's 80%. Sol was unchanged. This is just a limited promo for OpenRouter non-BYOK.
I'm pretty sure tokens are priced to maximize revenue, not inference profit.
I always find it funny that Japanese pensioners are probably subsidizing my tokens.
What do you mean?
SoftBank is one of the largest investors in openAI having contributed more than 30B and pledged another 30.
This is the opening salvos of an all out token price war.
With models a commodity at this point there isn’t much leverage for the big labs to keep their pricing anywhere near where it’s at. And that’s at the worst possible time as they need to be dramatically raising prices to have a viable business model.
Expect pricing to rapidly fall towards the underlying cost of compute and as players get really desperate we’ll likely see inference at less than the cost of compute as the market starts to rationalize and squeeze out weaker players who’s only play left will be to be the cheapest option in town.
The AI bubble is just waiting for the first player to scream mercy and cut capex as they simply can’t afford to throw more cash on the burning pile. That will be the trigger that implodes this bubble.
Since it's only discounted on the standard "OpenAI," non-ZDR route (old pricing on Azure), I'm guessing a lot of users won't see this benefit? Since a lot of users enable a global "ZDR-only" toggle on OR
It would seem that getting lots of data is exactly the reason to discount this.
OpenAI says they don't use any API data for training.
(there's probably going to be a reply about 'but how can you trust them'; I'm just stating what they say)
Unless it's in your contract, they will use the data. They might not be using it now, but they will eventually.
Does OpenRouter say the same?
dang, to avoid confusion from the title perhaps this should be edited to: “OpenRouter temporarily cutting GPT-5.6 Sol pricing by 50%”
That suggests this is being funded by OpenRouter, and there's no indication this is the case (and I doubt OpenRouter can afford it; why would they anyway).
Price wars did wonders for many businesses, like the bike sharing industry in China.
Overgrown datacenters or mounds of GPUs dumped into the harbour next ?
I would in such a scenario expect the GPUs to be dumped to industrial breakers who would send them to China for refurbishment and repackaging before being sold again on Amazon, AliExpress, and Taobao as last gen gaming cards from weird brands and specs.
This is what happened after the great crypto GPU dumping.
Honestly can't wait for that to happen, same with memory, drives, etc. There's going to be a massive amount of server pulls hitting the market.
The e-waste recyclers are pretty low on the pecking order, as the creditors will be first to strip these places for assets as Leopold Aschenbrenner discovered. =3
Yeah, I ment it as a joke - I agree with you. Watched the Gamers Nexus GPU investigation recently, where they were shown how a chinese soldering shop can transplant GPU chips to a new board, including memory chip reuse.
Hopefully we can look forward to all that useless datacenter AI crap gets repurposed in a similar manner into something actually useful for users.
I wouldnt be so hopeful because they dont use commodity hardware afaik How are you going to use a h100 at home?
with the same tricks gamers have always used? Modded drivers, undervolting, etc.?
Exactly + you can harvest the HBM memory on chip level & resolder on custom boards usable in PCs as VRAM (with scavenged GPU) or as normal RAM sticks.
I would happily buy up a load of datacenter GPU's at deep discount
Won't have to wait very long... as they are eating their own already.
https://www.youtube.com/watch?v=rE75WvOtcu8
The Shrek movie market correction correlation may be due again in July 2027. =3
Mountains of GPUs next to the ET games in the landfill.
Well one person can use at most one bicycle at a time.
One person can use as many GPUs as they want.
I like to see this. I still prefer Fable (marginally) but my last big task was 100% Codex using Sol max (re-sizing my AWS infrastructure using CDK) and it did a very good job. No complaints, I could use this model happily to do what I need to get done.
If this nudges Anthropic to give me more Fable usage, that's even better.
> my last big task was 100% Codex using Sol max (re-sizing my AWS infrastructure using CDK)
Fwiw, you could do this with any small or medium model, and it's easier with the aws-docs mcp. AWS is pretty stable, well documented, and programmatic, so most AI can figure out what it needs pretty quick
Hard to estimate what enabled the price cuts, Yet OpenAI is doing some magic work, especially recently.
Hard to estimate? Everyone knows the elephant in the room: capable open weight models.
Is this pricing change only for openrouter? I don't see official OpenAI info about this.
I was wondering the same, and it clearly says : 50% off, aka a sale, not normal price cut.
I don't get this thread.... Really. Is it full of bots?
Does this mean less subscription credit usage as well?
No because i am still losing 50% of my weekly quota using sol on high.
It just looks like OpenRouter is doing a 50% sale on some models right now? Until OpenAI makes an announcement, I would assume no.
OpenRouter doesn't do sales, they charge a premium, which is a flat 5.5% taken out of your credits. If you see something cheap on OpenRouter, it's because that one provider lowered its price. (Actually, correction, they will take 0.5% off their fee if you allow them to train your content)
Here are all the providers giving discounts: https://openrouter.ai/collections/discounted-models
Another thing some people don't notice is flex pricing, which is way lower than default pricing, for slightly worse latency and reliability. Depends on the provider and model
Slightly unrelated: what's up with the "tps" value? Does GPT-5.6 Sol really deliver just 32 tokens/second?
My bet is that OpenRouter began steering GPT-5.6-sol users towards flex tier, which is already 50% off.
So this isn’t really a price cut. As to why, lots of possible reasons. Perhaps an agreement with OpenAI to help them drive up more diverse traffic priorities.
At this point the models are “good enough” and whoever wins long term is gonna be whoever is the cheapest.
That’s why Chinese models are gaining traction and it’ll be the only way for OpenAI or Anthropic to keep up.
I'm loving this race to the bottom.
I'm not having that experience. So far each major model update has been at least slightly better than the last, in ways I've found useful. Can't say it's perfect, or able to do exactly what I want without a decent amount of instruction/implementation/docs, but it's been useful enough to keep paying for it.
GP means race to the bottom in price not quality.
Oh no the models are absolutely getting better, I'm just amazed that only 6 months ago I was using gpt-5.3-codex, and now I can use gpt-5.6-luna for similar results at like 1/15th the cost. Now 5.6-sol is being slashed by 50%? Amazing.
same, I have my popcorn - https://s-1.vercel.app/posts/why-openrouter-can-be-the-next-...
Is this motivated by the value of the thinking traces gleaned from the traffic?
The thinking traces are server-side, not exposed
They can’t decrypt the thinking traces.
You can train a LLM to inverse summarised thinking into thinking text. It’s not perfect, but it gets you maybe 80% of the quality with proper techniques.
Paper: https://arxiv.org/abs/2603.07267
FWIW, there’s not that much value protected here anyway IMHO, and even raw thinking text can lie (as shown by Anthropic’s amazing research), so for legitimate interpretability research it’s limited.
Scaling frontier performance hasn’t been SFT-bounded for a while now; it’s now basically how much you can scale RL rollouts.
The title looks to be misleading, since this price cut is limited to OpenRouter. It does not apply for the native OpenAI price listed at https://developers.openai.com/api/docs/models/gpt-5.6-sol
Yes, and it's only for a month. This is an ad.
Which raises the question - who is subsidizing this, and why?
Possibly OAI? If you have OAI tokens you are a captive audience. If you have OpenRouter you are bidding on a free market.
OpenRouter attributes this promotion to OpenAI https://x.com/OpenRouter/status/2089416739398254662
Why though, to AB test/see the impact of a price cut on a platform with multiple competitors?
Is this captive audience not going to switch providers for a 50% discount? Especially when the effort is simply swapping one URL for another?
OpenRouter is likely just leveraging Codex subscriptions.
Wouldn’t that be against TOS?
It might be through a level of indirection via an intermediate provider, offloading the TOS issue to the intermediate provider who couldn't care less.
Right? Should we switch from direct OpenAI API integration to OpenRouter?
What's the incentive here?
Open Responses API doesn't appear to support state management (yet)
OpenRouter and the Vercel AI Gateway.
So yes, presumably a very small share of their total traffic.
I saw this for Luna and then looked at the uptime and it said 85%. My interpretation is that this is just a gimmick where they serve the OpenAI flex tier at the same discount OpenAI provides for flex and then fall back to azure
Has anyone had mixed experience running Ultra with and without /goal? I come back to it after 8 hours to find it got stuck navel gazing imagined and Byzantine errors.
No ZDR. No dice.
Even at these prices, switching from subsidized subscriptions to the API just isn't worth it. Not even close.
In my experience, "Sol" stands for "Stupid overengineering LLM". I’ve tried it at low/medium/high/xhigh effort levels and after a while I always end up to regretting my switch from Opus/Fable.
Reading the comments in this thread, i honestly dont get it. 5.6-sol has felt like a regression in capability. In fact, every model since 5.3-codex has been a regression from OpenAI. I just find 5.6-Sol over engineers problems, takes absolutely ages to solve basic problems....
At this point, I'm considering going back to cursor over codex due to the ability to get more control over what model I use since there is clearly a heap of user preference and having frontier providers constantly shift the goal post with "State of the Art" is complete non-sense.
It depends on what effort you're using etc. As an example [1] of what codex is capable of, here's hugo (written in golang) ported to TypeScript - and then a TypeScript to Rust transpiler which converts arbitrary TypeScript into Rust.
The TypeScript code which was transpiled into Rust (and is compatible with most hugo templates) runs faster than the original hugo.
[1]: https://github.com/tsoniclang/tsonic-examples/tree/main/rust...
The transpiler is still WIP, but the fact that it can do this says a lot of about how far LLMs have come.
I find all effort levels of sol are the same in terms of amount of hallucinated unnecessary changes. Luna is much better all round on xhigh but my point still stands, every release of these new models is not an upgrade, its re-learning how to work with it.
Its like rehiring an employee every few months then training them up. Its honestly tiring and cant stay like this.
Opus has the same problem too…
That's not been my experience. My prompting methods haven't changed much between recent GPT releases. I do put a lot of effort into building tooling and tests around a project, so the LLM output is converging around it.
Can you give examples of the tooling and tests?
its over engineered problem solver ???? well because its a designed to do that
if you want to solve basic problem then use Luna
I mean it added additional changes when it doesnt need to. Its basically hallucinating changes it thinks it needs to make regardless of effort levels i try.
These threads seem to have become exceedingly vibes-based. Yes, something may now be cheaper or more expensive or whatever, but there is no way to objectively measure quality (except for "trust me bro" benchmarks). So the discourse is people saying that for them, this or that model was better - which is a very low value data point.
> These threads seem to have become exceedingly vibes-based.
If you remember programming language discussions, they are exactly like this.
Software development is still in the leeches and bloodlettings phase.
Yes, programming language discussions can be vibes-based and therefore low value too, but sometimes the more concrete aspects of the languages at hand, e.g. language features and tradeoffs are discussed. That is something that I'm not seeing in equivalent AI discussions.. there is a lot of how a particular AI model "feels" to interact with.
They are really getting desperate. The bubble is coming closer to an end here.
Does OpenRouter eat this cost to get their hands on a copy of the conversations people are using with the model?
No, this is OpenAI doing the discount, not Openrouter by themselves. OpenAI is crushing it with their 5.6 models, and they probably decided there was no better time to grab as much market share as possible.
I don’t understand this at all. They have never been profitable yet. How is this helping them? When it be more likely the case that not enough, people are using it as the prices they established already? So now they have to lower the prices?
You lower prices for marketshare. Fable became a mythological model to leadership because they were the first story of ai escaping and hacking another company. The it's so dangerous the public can't use it narrative is sticky so OpenAI is showing off its model so as many eyeballs as possible. We're in the samples in the supermarket phase.
They're prepping to IPO. They want top level metrics like usage they can use to pump investors, not nonsense like profitability.
OpenRouter offers 1% discount to save your conversations, explicitly opt-in.
Anything else they don't save it. Even if they tell you the model provider saves your data for training.
For context, Stripe has just acquired OpenRouter for >$7B.
I’d bet that explains this move!
Why would the potential acquisition have anything to do with this? They do discounts all the time on various models. Luna was 50% off last week..
Not sure as OpenAI models (Sol, Luna,..) are also discounted on the Vercel AI Gateway rn. My bet is on OpenAI trying to drive more enterprise customers to their models through API.
How can they do this? Are they subsidizing it out of pocket?
Terra is also a fantastic model.
GPT-5.6 sol starting to be a real workhorse at this price point
It shocks me how little people seem to care that they are supporting an evil Zionist lizard man who molested his sister and is happy supporting trump. Doesn’t even come up in the conversation here. I don’t really care if sol is a bit better, I still make decisions on more than that.
Is the HN community just too online and sucked in to the musk mind manipulation vortex? Or what is going on? Why does nobody seem to care?
Can we get it for the reduced rate direct from OpenAI though?
Do other people find 5.6 to be worse at most simple tasks and frequently over complicate things?
I asked it to write a user todo and it turned out a four page essay. I gave the same task to 5.4 and got the small list of checkboxes I expected.
I've found it to be great for planning code changes (or new projects). I use the superpowers plug-in which I think guides the planning.
Then I switch models (to luna) before implementation. I find this combo nearly always does what I want.
I also use a skill called ponytail, its goal is to keep things terse and edits small. It may have contributed to the successes above.
I like that skills are easy to try out, too.
I stopped using superpowers because it wanted to turn every tiny bug fix into a $37MM DOD project. I got effective results but it took ages. I may try again - I need to find a good way to run different profiles in my harness so I can easily shut it off. The default planning workflow in OMP is pretty good though.
I agree Luna is great for task execution, either as a sub-agent with Sol planning and coordinating or if the task is well defined and straightforward, but there are lots of models now that you can say that about.
I have the same setup you have, love it!
You would probably get better results with Luna for the real simple tasks, or Sol with low thinking effort.
I find that I get exactly the effort that I asked for, which is pretty nice. The other side of that coin is that these are the least lazy models I’ve used so far. They will go on elaborate tangents to complete the task when I want them to.
any good best practice for effort selection on claude. I always use high as default.
I've found it's worse for simple tasks too, and I have to give it stricter guidelines, and sometimes it doesn't follow the same patterns I've grown to expect. I've found using 5.6 (sol) is good for diagnosing issues though, especially in terms of optimization of some given path
Yep. I have not yet had a single good experience with Sol or the 5.6 models on a variety of harnesses and configurations. It overthinks, overcomplicates and often makes my code into an unmaintainable sludge. It'll usually take 5+ turns of steering to get it in the right direction.
It's your responsibility to set an appropriate level of Thinking. For simple tasks, I use the instant model. As an approximation, the choice is proportional to the amount of time I want it spending on the task. Also, you can always ask it to respond succinctly.
Absolutely not
I’ve used Claude exclusively for the past few months
Was excited when Sol came out a few weeks ago and loaded it up
I made the mistake of treating it as if it were Claude - I’d assumed they were close enough in ability and treated them that way
Well, turns out my instruction sets for Claude are 100% too complicated for Sol
Sol made the stupidest assumptions, constantly did things that it wasn’t asked to do and always approached code in what I considered a weird way - I had redo a lot of my prompts to get it anywhere close
Now, did it do good work?
Yes, on occasion. But with LLMs and coding, consistency is the name of the game. Constantly having to correct the LLM and constantly feeling paranoid that it won’t listen makes for an exhausting session
Maybe if you “came up” in the codex world you’re more fluent with it, but sticking with Claude for now
"rewrite unreal engine in rust, make no mistakes"
Why are you being downvoted, is this post an ad or something.
Probably because it's PEBCAK.
Hard to take offense from someone who uses pebkac seriously
Kudos to you though for being your authentic self so publicly
Oh hey, that's cheaper than Kimi K3! Amusing to see a SOTA OpenAI model be cheaper than a Chinese open weight model.
Fwiw I love K3 and use it as a daily driver. I haven't tried Sol, as I dislike OpenAI.