ValsAI 新基准测 GPT-6 Astra 推理档位性价比

Vals AI @ValsAI原推发布于 2026.09.29

在 X 上打开原推

When does more reasoning stop being useful and start becoming overhead? We ran Open AI and Anthropic’s best models: GPT-6 Astra and Claude Opus 5.5 at every reasoning level on our new benchmark, MysteryMechanism. Thinking longer delivers accuracy gains with the same model at as much as 20x the cost. Inspecting traces, the models use their effort level in very different ways.