Unverified50% confidenceBenchmarkExact time
Among the 14 benchmarks Meta itself published, Muse Wonder won 5, mainly in long-horizon coding, legal agent tasks, and Meta's internal memory tests; it lost 9, including ProgramBench (Opus leads 70.5 to 50.0).
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/12/2026
First Seen
Valid until: 10/26/2026
Sources
Related Entities
Related Claims
UnverifiedAmong the 14 benchmarks listed on Meta's model card, Muse Wonder (the API version) achieved 5 wins and 9 losses against Claude Opus 5.5.77% similarUnverifiedBenchmarks show that when the number of concurrent requests reaches 5, the throughput ratio of row-by-row GEMV over GEMM drops below 100% (90.7%), and GEMM regains the advantage.69% similarUnverifiedMeta's chart shows Muse Wonder beating Claude Opus 5.5 on long-horizon coding tasks: DeepSWE score 78.8 versus 74.2, SWE-Marathon score 53.1 versus 46.3.68% similarUnverifiedOn the Automation Bench test, the high-tier Sol scored higher than the highest-tier Cloud Ops 5, with a per-question cost of only 9% of its competitor's.65% similarUnverifiedOn ProgramBench, Muse Wonder scored 50.0 and Claude Opus 5.5 scored 70.5, with Opus ahead by 20.5 points, the largest gap among the 14 benchmarks.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1011214API
curl https://kongchang.com/api/v1/knowledge/claims/1011214MCP
get_claim(id=1011214)