godjan

·昨年·議論

The article doesn't mention one of the most complex benchmarks - ARC challenge. All models suck in it https://arcprize.org/leaderboard

but Gemini and Claude still suck much worse then ChatGPT models