Meta AI·released 2026-04-081 source
Muse Spark benchmark scores: 8 benchmarks tracked, leading 1. 63% of its scores are independently reproduced — the rest are self-reported or unverified.
Leads on 1 of 8 benchmarks·best result 88.89 on OTIS Mock AIME 2024-2025 (reproduced)·5 of 8 independently reproduced
Head-to-headMuse Spark vs Gemini 3 ProLeads on: FrontierMath-2025-02-28-Private
| Benchmark | Score | Evidence status | Measured |
|---|---|---|---|
| OTIS Mock AIME 2024-2025 | 88.89accuracy (%) | reproduced· optimizedT1 | 2026-04-08Epoch AI |
| GPQA diamond | 86.4accuracy (%) | reproduced· optimizedT1 | 2026-04-08Epoch AI |
| FrontierMath-2025-02-28-Private | 68.42accuracy (%) | reproduced· optimizedT1 | 2026-04-08Epoch AI |
| SimpleQA Verified | 66.3accuracy (%) | reproducedT1 | 2026-04-08Epoch AI |
| HLE | 37.56accuracy (%) | unverifiedT2 | 2026-04-08 |
| FrontierMath-Tier-4-2025-07-01-Private | 24.33accuracy (%) | reproducedT1 | 2026-04-08Epoch AI |
| ProofBench | 17accuracy (%) | unverifiedT1 | 2026-04-08https://www.vals.ai/benchmarks/proof_bench |
| CritPt | 11.33accuracy (%) | unverifiedT2 | 2026-04-08 |
Claims drawn from cited facts, not live model generation.
This model was developed by Meta AI and released in 2026.
2 cited facts
This model is scored on 8 tracked benchmarks and currently holds the top score on 1 of them. Because the results have been independently reproduced, the scores are backed by outside evaluation and can be trusted.
3 cited facts
With an average gap of 18.11 points, the model trails the leader by a wide margin, placing it well behind the front-of-pack on its benchmarks. It holds the top score on 1 benchmark, indicating genuine leadership in that single category, but the overall average gap still puts it far from overall frontier status.
2 cited facts
1 facts cross-checked across data sources: 1 single-source.
Source facts, citations, and refresh stamp for this record.
Sources: epoch_benchmarks
Muse Spark is an AI model developed by Meta AI, released 2026-04-08. Its benchmark record below tags every score as independently reproduced, self-reported, or unverified.
Muse Spark has recorded scores on 8 benchmarks, each shown with its evidence status.
Muse Spark has recorded scores on 8 benchmarks — GPQA diamond, OTIS Mock AIME 2024-2025, FrontierMath-2025-02-28-Private, SimpleQA Verified, CritPt, FrontierMath-Tier-4-2025-07-01-Private, and 2 more. The full table above shows each score with its evidence status.
5 of 8 recorded scores (63%) are independently reproduced rather than self-reported by the lab.
Muse Spark has 8 tracked claims: 5 independently reproduced, 3 unverified.