Expired50% confidenceBenchmarkExact time
In Harvey LAB-AA legal analysis, Muse Wonder achieved 13.1%, more than three times Claude Opus 5.5's 4.2%; in Vals Finance Agent V2 it ranked first with 59.1%.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/13/2026
First Seen
This claim has expired. It is kept for the record and no longer describes the current state.
Valid until: 10/27/2026(expired)
Sources
Related Entities
Related Claims
UnverifiedArtificial Analysis gives Muse Wonder an intelligence index of 52, ranking 9th out of 219 models; Claude Opus 5.5 is 58, Astra is 53, and Wonder Full is 56.61% similarUnverifiedThe GLM-Legal model launched by Z.ai achieved a 19.2% all-pass rate on the Harvey LAB legal Agent benchmark, surpassing Claude Fable 5 and Kimi K3.58% similarUnverifiedIn the coding agent test, Sol output more than half fewer tokens and cost approximately 1/3 of Faber-5.57% similarUnverifiedThe Comprehensive Intelligence Index consists of 9 independent evaluations including agentic work, coding, scientific reasoning, and general capability; Sol and Faber-5 are nearly on par with a difference of only 1 point, but task completion time was reduced by approximately 61% and cost was approximately half.56% similarUnverifiedSol scored 68.8% in the coding test, close to the previous-generation Fable 5's 69.9%, but with a per-task cost approximately 80% lower.56% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1020384API
curl https://kongchang.com/api/v1/knowledge/claims/1020384MCP
get_claim(id=1020384)