Vals AI
@ValsAI
参加 March 2024
277 フォロー中 22.8K ファン
On Terminal-Bench 2.1, we deterministically screened 1,602 trajectories for solution code pulled verbatim from the internet. GPT-5.6 Terra led at 4.5%, with Gemini 3.8 Flash close behind at 2.6%. Two of the newest releases, again near the top.
もっと見る
