gemini-4IMAGE
KnotBench 八项任务横评：Gemini 4 Argon 对比 GPT-6 Astra 等五模型
作者用自建 KnotBench（跑代码、找 bug、写最短程序、自评分数、承认无解等 8 项）测试 Claude Fable 5.1、GPT 6 Astra、GPT 6.1 Sol、Grok 4.7 和 Gemini 4 Argon。
Athrix ☄️
@athrix_codes17 天前收录原推发布于 2026.10.18
Athrix ☄️ @athrix_codes原推发布于 2026.10.18
I benchmarked 5 frontier models on KnotBench Models - Claude Fable 5.1 - GPT 6 Astra - GPT 6.1 Sol - Grok 4.7 - Gemini 4 Argon What is KnotBench? > 8 tasks: run code, find a bug, write the shortest program, guess your own score, admit when there's no answer
x.com/athrix_codes/status/2111842775457234992采集于 2026.10.19 09:58