Unverified50% confidenceBenchmarkExact time
在Terminal Bench 2.1基准测试上，GPT-5.6 Sol的Ultra模式接近92%，超越竞品的88%表现
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/20/2026
First Seen
Valid until: 10/18/2026
Sources
Related Claims
UnverifiedTerminal Bench 2.1测试中GPT-5.6 Ultra得分接近92%，Fable 5为88%，考虑误差范围基本持平79% similarUnverified在SWE-Bench Pro编程评测上，Claude Fable 5以80%的成绩领先GPT-5.6 Sol的64.6%79% similarUnverified在Artificial Analysis的Omniscience幻觉指标上，GPT-5.6 Sol Max高达89%，高于Claude Fable的55%和GLM 5.2的28%79% similarUnverified普通版 Sol 在 Terminal Bench 2.1 基准测试中获得 88.8%，开启 Ultra 模式后提升至 91.9%75% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/573057API
curl https://kongchang.com/api/v1/knowledge/claims/573057MCP
get_claim(id=573057)