Muse Spark Open
Category scores
Scores per capability area with the model's rank among ranked models in that category.
- 1Multimodal83.6
- 2Reasoning78.8
- 3Knowledge63.3
Benchmark results (9)
As reported in system cards, papers and public evaluations. Position compares this model with every other model reporting the same benchmark.
Agentic & tool use
| MCP Atlas | 89.2 | #1 of 42 |
| JobBench | 61.8 | #6 of 34 |
| HLE w/ tools | 61.0 | #6 of 27 |
| Toolathlon-Verified | 72.8 | #14 of 22 |
Multimodal
| SimpleVQA | 73.6 | #4 of 17 |
| ZeroBench | 50.0 | #1 of 8 |
| BabyVision | 80.2 | #2 of 8 |
Coding
| DeepSWE | 60.0 | #35 of 50 |
| SWE-Atlas Codebase QnA | 58.0 | #3 of 4 |
Pricing by source
USD per 1M tokens.
| Source | Input | Output | Cached in |
|---|---|---|---|
| List price (OpenRouter) | $1.25 | $4.25 | — |
| OpenRouter meta/muse-spark-open | $1.25 | $4.25 | $0.15 |
Meta publishes downloadable Muse Model License weights and names hosted API partners (Baseten, Fireworks AI, Groq, Modal, Nebius, OpenRouter, Together AI). Neither the launch post nor the model card publishes a per-token price, and the Hugging Face page lists no deployed inference provider at the time of review. Partner prices are not copied here.
Arena ratings
Human-preference ratings from LMArena. Best-rated variant shown.
| WebDev (coding)(Max) | 1528 | #52 · 1.4K votes |
Open weights
From the Hugging Face repository.
- Parameters
- 822B
- Licence
- other
- Downloads (30 days)
- 2.6K
- Repository
- meta-models/Muse-Spark-Open-816B-A42B ↗
Speed
Measured by us through OpenRouter (median of trials).
Not measured yet.
Related models
Similar score
- Gemini 3.6 Flash62.8 · $1.50
- Qwen3.8-Flash-Next62.9 · —
- Qwen3.7 Max62.6 · $2.21
- DeepSeek V4 Pro 081363.2 · $0.99
- Claude Opus 4.662.3 · $10.00