WHAT NEW IN X11.06 FRI

AI collections from X · 23 collections · 69,408 posts

anthropicIMAGE

Claude Opus 5.5 beats GPT-6 Astra on most benchmarks at lower cost

The post lists Opus 5.5 leading Astra on Terminal-Bench, FrontierCode, knowledge-work Elo and computer use at about 40% lower cost, with Astra only slightly ahead on business automation and research agents.

SuSu_酥酥👅
@NFT_ChenArchived 2026.09.23Originally posted 2026.09.23
Claude Opus 5.5 beats GPT-6 Astra on most benchmarks at lower cost
SuSu_酥酥👅 @NFT_ChenOriginally posted 2026.09.23
Open original on X

🚨炸裂!Claude Opus 5.5 来了!碾压GPT-6 Astra!日常旗舰直接摸到 Fable 5.1 的天花板,还更便宜、更快! 🔸智能体编程三连冠:Terminal-Bench 66.4%(Fable 55.8 / Astra 57.9 / Sol 37.3),FrontierCode 54.4%,CursorBench 57.8% 
🔸知识工作 1846 Elo,全面压过 Fable、Astra、Sol 
🔸电脑使用 81.8%、图表识别 89.0% 
🔸比 Opus 5 典型成本低约 40%,$4/$20,缓存读取 $0.20,输出快 30%+ Astra 只在业务自动化、科研智能体上略胜;Sol 全线落后。68 万行迁移一天搞定,性价比这一枪打穿了。 #Claude #Opus55 #Anthropic #GPT6Astra #AICoding

This post quotes another post: x.com/nft_chen/status/2102435847836758453

x.com/nft_chen/status/2102437769218003227Archived 2026.09.23 10:28