Meta model product
Muse Spark is the first model in the Muse family developed by Meta Superintelligence Labs.
Updated Aug 12, 2026. Default version: Muse Spark
Technical details for the model's default version.
This profile uses the model's current scored version. Arena ratings and prices are shown separately.
Benchmark scores for Muse Spark.
| FrontierScience Research | 38.3% | 01 | 4 | 100.0% | C | |
| HealthBench Hard | 42.8% | 01 | 9 | 100.0% | C | |
| IPhO 2025 | 82.6% | 01 | 2 | 100.0% | C | |
| MedXpertQA | 78.4% | 01 | 12 | 100.0% | C | |
| ZEROBench | 0.33 points | 02 | 9 | 87.5% | C | |
| LiveCodeBench Pro | 0.8 points | 03 | 4 | 33.3% | C | |
| ScreenSpot Pro | 84.1% | 04 | 25 | 87.5% | C | |
| SimpleVQA | 0.713 points | 04 | 13 | 75.0% | C | |
| Humanity's Last Exam | 58.4% | 05 | 93 | 95.7% | C | |
| DeepSearchQA | 74.8% | 08 | 9 | 12.5% | C | |
| ERQA | 64.7% | 08 | 23 | 68.2% | C | |
| ARC-AGI v2 | 42.5% | 09 | 16 | 46.7% | C | |
| CharXiv-R | 86.4% | 09 | 48 | 83.0% | C | |
| MMMU-Pro | 80.4% | 13 | 66 | 81.5% | C | |
| Tau2 Telecom | 91.5% | 16 | 35 | 55.9% | C | |
| Terminal-Bench 2.0 | 59.0% | 24 | 49 | 52.1% | C | |
| SWE-Bench Verified | 77.4% | 25 | 105 | 76.9% | C | |
| GPQA | 89.5% | 26 | 234 | 89.3% | C | |
| SWE-Bench Pro | 52.4% | 40 | 45 | 11.4% | C |
Preference and agent-evaluation results for the default version.
| vision style control | overall | 08 | 1294.2 | 5,576 | N/A | |
| text style control | overall | 11 | 1488.0 | 13,602 | N/A | |
| vision | overall | 11 | 1306.7 | 5,576 | N/A | |
| document style control | overall | 18 | 1466.3 | 1,078 | N/A | |
| text | overall | 18 | 1473.1 | 13,602 | N/A | |
| document | overall | 24 | 1443.2 | 1,078 | N/A | |
| text factuality | overall | 25 | 1466.2 | 13,585 | N/A |
Provider-specific output speed and catalog latency for Muse Spark. Runtime does not affect the capability score.
No provider-specific speed or latency record is linked to the default version yet.
Output Speed is generated output tokens received per second. Catalog latency is reported separately from observed provider TTFT.
Official vendor API pricing appears first, followed by individual provider offers.
The default version has no current input or output token prices.
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
Available versions of this model. The score column identifies the version used in the overall ranking.
| Muse Spark | 73.2 | N/A | N/A | N/A | No | Proprietary |
Key information about Muse Spark and its available data.
Muse Spark is the first model in the Muse family developed by Meta Superintelligence Labs. It is a natively multimodal reasoning model with support for tool-use, visual chain of thought, and multi-agent orchestration. It features a Contemplating mode that orchestrates multiple agents reasoning in parallel.
It demonstrates competitive performance in multimodal perception, reasoning, health, and agentic tasks, with Contemplating mode achieving 58% on Humanity's Last Exam and 38% on FrontierScience Research.
Data as of 2026-08-11.
Open a comparison with the three ranked models immediately above and below this model.
Recommendations prioritize the same model type and family, then the closest LLMBoard score.
Common questions about Muse Spark.
Muse Spark's default version was released on Apr 8, 2026.
No official standard PAYG price is currently available for Muse Spark.
Muse Spark was created by Meta.
A context window is not available for the default version.
No. The default version is not marked as having publicly available weights.
No provider offering is currently linked to the default version.
Nearby ranked alternatives include Kimi K2.6, DeepSeek-V4-Pro, Qwen3.7 Plus.