I think a major reason is that China has 1.4 billion people, and Chinese developers can use much cheaper AI models like DeepSeek and Qwen. As a Chinese person, I'm glad to see more of these affordable and powerful models, and I don't like closed-source ones that much.
I think a major reason is that China has 1.4 billion people, and Chinese developers can use much cheaper AI models like DeepSeek and Qwen. As a Chinese person, I'm glad to see more of these affordable and powerful models, and I don't like closed-source ones that much.
But if closed-source models are cheaper, I won’t hesitate to support them either, hahaha
That may is true. But I can't use closed models on my own pc.
for relative comparison, google reported they process 3.2Q tokens / month this last i/o
https://chatforest.com/builders-log/google-3-2-quadrillion-t...
OAI reported 15B / min around the same time, which if my math is accurate, is .67Q tokens / month
https://www.feedough.com/openai-statistics/