2026-11-09 · America/Los_Angeles · 社区动态 · #2

Is ArtificialAnalysis biased against open weight models?

It seems like whenever open weight models come close to proprietary US models, ArtificialAnalysis conveniently "re-calibrates" their benchmarks which then shoves open weight models down the toilet. Then they advertise on X/Twitter this "intelligence index" which presents open weight models as dog shit compared to Claude or GPT (all while ignoring the actual intelligence of models like DeepSeek or GLM in tasks that actually matter, which is coding vs telecom bullshit). Any business man looking at these charts would conclude "open weight models are shit, Claude is where its at, I should get a…

热度 48.0 / 100；排名与评分保留该期记录。