Unverified50% confidenceOpinionExact time
Jane Street评价Astra在内部编程基准上达到SOTA并超过Fable,但在交易直觉上只是明显进步(Fable 5.1交易直觉略胜)
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/4/2026
First Seen
Valid until: 1/2/2027
Sources
Related Entities
Related Claims
Unverified在Frontier Code编程测试上,Opus 5.5用默认档赢了Astra,单任务成本只有对手的五分之一58% similarUnverified在该次OZ世界还原测试中,Astra整体观感最佳,Opus 5.5细节最多,Sol 6.1表现明显掉队58% similarUnverified在Frontier Code基准等长周期编程任务上,Opus 5.5可能略胜Astra58% similarUnverified在作者自制的代码审查基准中,Astra位居榜首,Grok 4.7紧随其后,Fable 5.1只提5条建议但质量很高57% similarUnverifiedSol 在编程智能体指数全推理模式下拿到 80 分,刷新 SOTA,比 Faber-5 高 2.8 分53% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/972630API
curl https://kongchang.com/api/v1/knowledge/claims/972630MCP
get_claim(id=972630)