Meta
availableShows if the model has enough results for an index.Muse Spark
Muse Spark is a reasoning model from Meta. 31 benchmarks count toward its score, in 7 categories.
IndexOverall score out of 100.61.6 ±3.4
CoverageShare of the index weight with results.95%
SpeedOutput tokens per second.—
Input / 1MUS dollars per 1M input tokens.N/A
Output / 1MUS dollars per 1M output tokens.N/A
ContextMaximum tokens in one request.262K
EloLMArena rating and rank.1473 (#22)
The index is a score out of 100. The ± range shows how much it can change.
13,565 votes. Elo shows what people prefer. It does not change the score.
CapabilitiesScore per category, out of 100.
Out of 100Results
31 counted| BenchmarkThe test name. | CategoryThe capability that the test measures. | ResultThe score from the publisher. | IndexThis result as a score out of 100. | RunThe settings of the run. | DateDate of the result. | Published byThe source of the result. |
|---|---|---|---|---|---|---|
| AIME | Math | 96.9% | 58.9 | — | 16 Apr 2026 | Vals AI |
| τ²-Bench Tool-Agent-User Evaluation | Agentic | 91.5% | 64.5 | — | — | Victor Barres et al. |
| GPQA diamond | Knowledge | 89.8% | 61.1 | — | — | Epoch AI |
| GPQA Diamond | Knowledge | 89.6% | 61.0 | — | 1 Sept 2026 | Vals AI |
| GPQA Diamond | Knowledge | 89.5% | 60.9 | — | — | David Rein et al. |
| OTIS Mock AIME 2024-2025 | Math | 88.9% | 61.2 | — | — | Epoch AI |
| Artificial Analysis GPQA Diamond | Knowledge | 88.4% | 59.1 | — | — | Artificial Analysis |
| MMMU Pro | Multimodal | 87.4% | 69.6 | — | 1 Sept 2026 | Vals AI |
| MMLU Pro | Knowledge | 87.3% | 58.1 | — | 1 Sept 2026 | Vals AI |
| CharXiv Reasoning | Multimodal | 86.4% | 64.5 | — | — | CharXiv authors |
| ScreenSpot Pro | Multimodal | 84.1% | 66.1 | — | — | Kaixin Li et al. |
| Artificial Analysis MMMU-Pro | Multimodal | 80.5% | 65.5 | — | — | Artificial Analysis |
| Massive Multi-discipline Multimodal Understanding Pro | Multimodal | 80.4% | 58.2 | — | — | MMMU-Pro authors |
| LiveCodeBench Pro | Coding | 80.0% | — | — | — | LiveCodeBench Pro authors |
| MedXpertQA Multimodal | Multimodal | 78.4% | — | — | — | Meta AI |
| Artificial Analysis Long Context Reasoning | Reasoning | 78.0% | 62.2 | — | — | Artificial Analysis |
| Software Engineering Benchmark Verified | Coding | 77.4% | 59.6 | — | — | Carlos E. Jimenez et al. |
| Artificial Analysis IFBench | Instruction | 75.9% | 66.1 | — | — | Artificial Analysis |
| DeepSearchQA | Agentic | 74.8% | 57.4 | — | — | Meta AI |
| SWE-bench | Coding | 74.4% | 57.2 | — | 1 Sept 2026 | Vals AI |
| SimpleVQA | Multimodal | 71.3% | 63.8 | — | — | Z.AI |
| ERQA | Multimodal | 64.7% | 59.4 | — | — | Qwen |
| Claw-Eval | Agentic | 63.8% | 58.7 | — | — | Bowen Ye et al. |
| Terminal-Bench 2.0 | Agentic | 59.6% | 65.2 | — | 4 Jun 2026 | Vals AI |
| Artificial Analysis Coding Index | Coding | 58.6% | 60.3 | — | — | Artificial Analysis |
| MedXpertQA Text | Knowledge | 52.6% | — | — | — | Meta AI |
| SWE-bench Pro | Coding | 52.4% | 54.7 | — | — | Xiang Deng et al. |
| Humanity's Last Exam | Knowledge | 50.4% | 71.5 | — | — | Center for AI Safety et al. |
| Artificial Analysis Omniscience Accuracy | Knowledge | 49.6% | 75.2 | — | — | Artificial Analysis |
| CyberGym | Agentic | 43.5% | 46.0 | — | — | Zhun Wang et al. |
| Humanity's Last Exam without tools | Knowledge | 42.8% | 65.0 | — | — | OpenAI |
| HealthBench Hard | Knowledge | 42.8% | 86.4 | — | — | Meta AI |
| Artificial Analysis Humanity's Last Exam | Knowledge | 40.7% | 68.7 | — | — | Artificial Analysis |
| FrontierMath-2025-02-28-Private | Math | 39.0% | 67.0 | — | — | Epoch AI |
| ZeroBench | Multimodal | 33.0% | — | — | — | Meta AI |
| Artificial Analysis Intelligence Index | Knowledge | 31.3% | 61.5 | — | — | Artificial Analysis |
| GDPval-AA normalized | Agentic | 24.3% | 54.2 | — | — | Artificial Analysis |
| Vibe Code Bench v1.1 | Coding | 19.7% | 50.4 | OpenHands | 21 Sept 2026 | Vals AI |
| FrontierMath-Tier-4-2025-07-01-Private | Math | 14.6% | 59.3 | — | — | Epoch AI |
| Critical Physics Tasks | Reasoning | 11.3% | 62.1 | — | — | Artificial Analysis |
31 benchmarks count, from 36 of 40 results. A grey row does not count. Too few models took that benchmark.