生产项目实测：Gemini 4 Argon 59 分 vs Opus 5.5 77 分

bidyut @bidyut_io原推发布于 2026.10.14

在 X 上打开原推

gemini 4 argon vs opus 5.5. wasn't expecting this 👀 tested both on a real feature in my production python backend project. opus 5.5 (extra high): 77/100 gemini 4 argon (high): 59/100 gpt 6 astra judged both implementation plans. argon also agreed with the review. early test, so more testing needed. but argon compacted context surprisingly fast. not sure what context window antigravity is using, but it didn't feel like 1M tokens. also, we need an extra high reasoning level.