New todaygpt-6IMAGE

FrontierCode 2.0: Opus 5.5 tops, GPT-6 Astra second at 42.8%

Cognition's FrontierCode 2.0, which tests real-world engineering judgment, tops out at 45.5% with Opus 5.5, followed by GPT-6 Astra at 42.8%.

SuSu_酥酥👅

@NFT_ChenArchived 9 hours agoOriginally posted 2026.10.28

FrontierCode 2.0: Opus 5.5 tops, GPT-6 Astra second at 42.8%

SuSu_酥酥👅 @NFT_ChenOriginally posted 2026.10.28

Open original on X

🔥最新硬核基准！Cognition发布FrontierCode 2.0，测的是「你会不会真merge、该不该merge」故意模糊需求，考察真实工程判断力！最强也只有45.5%，远未饱和： 🔹头部：Opus 5.5 45.5%｜GPT-6 Astra 42.8%  🔹中高端：GPT-6.1 Sol 39%仅$1.03  🔹性价比：DeepSeek V4.1 Flash 19.3%｜Kimi K3 / GLM 约17.7%  真正接近生产场景的评测。榜单：https://cognition.com/frontiercode #AI #大模型 #编程 #FrontierCode #DeepSeek #Qwen #开源模型 #梁文峰 #杨植麟

This post quotes another post: x.com/cognition/status/2115132488775524793

x.com/nft_chen/status/2115164670944584078Archived 2026.10.28 10:11