Hacker News Reader

DeepSeek hasn't been SotA in at least 12 calendar months, which might as well be a decade in LLM years

What about Kimi and GLM?

These are well behind the general state of the art (1yr or so), though they're arguably the best openly-available models.

According to artificial analysis ranking, GLM-5 is at #4 after Claude Opus 4.5, GPT-5.2-xhigh and Claude Opus 4.6 .

Idk man, GLM 5 in my tests matches opus 4.5 which is what, two months old?

4.5 was never sota