OpenAI GPT-OSS models use MXFP4 to cut inference costs

(theregister.com)

8 points | by rntn 3 days ago ago