linzhangrun

There are too many commoditized models to count: GLM5.3 Flash, Kimi K2.8, Mimo V2.6, MiniMax M3.1...

Just like AI-related projects on GitHub always get an entirely different order of magnitude of stars. This is what people are focused on, outshines anything else.

Actually, the mind should only exist in the forebrain:0

How surprising.

The speed of light, wave-particle duality, and so on, always make me feel that the universe also has a "computing power"

They say v4.1flash is so strong that they'll route API calls to v4pro to v4.1flash, lol

super fast true

update: Refresh and it won't appear again

Obviously a Fields Medal–level achievement... Even the most optimistic person would not have expected it to happen so quickly six months ago

Under the same computing power, the improvement in LLM intelligence and the improvement in the upper limit of LLM intelligence are equally astonishing. At least in coding, the best local models that can run smoothly on a DGX Spark are now less than one year behind the strongest SOTA models in capability (I measured around 60 tok/s).