Expired60% confidenceBenchmarkExact time
GPT-5.6 Sol in standard mode surpasses Claude Mythos 5 on Terminal Bench 2.1, and with Ultra Mode enabled, the score reaches 91.9%.
2
Sources
60%
Confidence
Medium-term (~90 days)
Relevance
7/12/2026
First Seen
Valid until: 10/10/2026(expired)
Sources
GPT-5.6 Sol Hands-On: How Multi-Agent Parallelism Is Reshaping AI Coding Workflows
bilibili地层世界7/11/2026
Related Claims
UnverifiedGPT-5.6 achieved a score of 91.9% on the Terminal-Bench 2.1 coding leaderboard with the Ultra configuration.80% similarVerifiedGPT-5.6 Sol achieves state of the art on the coding benchmark Terminal Bench.78% similarVerifiedSOL's benchmark score is nearly double that of the previous generation GPT-5.5, representing a generational leap, but still trails Fable 5 overall.77% similarUnverifiedGPT-5.6 Sol introduces two new reasoning levels, Max and Ultra.77% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/492456API
curl https://kongchang.com/api/v1/knowledge/claims/492456MCP
get_claim(id=492456)