All tasks / Models
Muse Spark Open (816B-A42B)
Meta's open-weights Muse Spark model, an 816B-parameter MoE with 42B active, under the Muse Model License.
- Meta
- Open weights
- Released 15 Oct 2026
Where it ranks
- CChat & assistantsModels · Local · 39 on Artificial Analysis Intelligence Index (secondary benchmark)
- CCodingModels · Local · 1,524 on LMArena WebDev / Code Arena (overall)
- AImage understandingModels · Local · 80.9% on Artificial Analysis MMMU-Pro (independent evaluation) (secondary benchmark)
- CWorkflow automationModels · Local · 49.7% on AutomationBench-AA (Artificial Analysis)
Facts
- Vendor
- Meta
- Family
- Muse Spark Open
- Released
- 15 Oct 2026
- Weights
- Open: download and run it yourself
- License
- Muse Model License
- Parameters
- 816B (42B active)
- GPU memory to run locally
- ~470 GB (4-bit ≈ 816 B × 0.57 GB + overhead ≈ 470 GB, estimate (multi-GPU datacenter); an NVFP4 checkpoint is published.)
- API price
- $1.25 in / $4.25 out per 1M tokens
- Access
- Free (open weights)
- Last verified
- 18 Oct 2026
Price from openrouter.ai, 18 Oct 2026.
Benchmark results
| Benchmark | Result | Entry on the leaderboard | As of | Source |
|---|---|---|---|---|
| Artificial Analysis Intelligence Index | 39 | Muse Spark Open (Xhigh)Raw value from the page's embedded model data; the site displays it rounded to an integer. | 18 Oct 2026 | artificialanalysis.ai |
| Artificial Analysis MMMU-Pro (independent evaluation) | 80.9% | Muse Spark Open (Xhigh)Raw fraction 0.8092 from the model data embedded in https://artificialanalysis.ai/leaderboards/models, converted to percent; rank 22 of 124 entries with an MMMU-Pro result; no last-updated date shown. | 19 Oct 2026 | artificialanalysis.ai |
| AutomationBench-AA (Artificial Analysis) | 49.7% | Muse Spark Open (Xhigh)AutomationBench-AA Score (share of each task's objectives completed, a task with any guardrail violation scoring zero); raw value from the page's embedded model data (displayed rounded); rank 26 of 32; Tasks Completed 20.85%, Objectives Completed 76.30%; no last-updated date shown, so as_of = retrieval date. | 19 Oct 2026 | artificialanalysis.ai |
| LMArena WebDev / Code Arena (overall) | 1,524 | Muse Spark Open (Max)From LMArena's official leaderboard-dataset (webdev = Code Arena, split latest, category overall, published 2026-10-18), which mirrors https://arena.ai/leaderboard/webdev. Rank 50; 1337 votes; 95% CI 1506.6-1541.3. | 18 Oct 2026 | huggingface.co |
| SWE-Bench Pro V2 (public set, full) | 95% | Muse Spark Open (mini-swe-agent) xhighSWE-Bench Pro V2 Full (642 tasks); row added 2026-10-16; 95% CI +/-1.43; harness in the variant name; read from the leaderboard page's embedded data. | 19 Oct 2026 | labs.scale.com |
Where you can use it
- AGroqRuntimes & gateways · Cloud
- SHugging Face Inference ProvidersRuntimes & gateways · Cloud
- SHuggingChatApps · Cloud
- SOllamaRuntimes & gateways · Local
- SPoeApps · Cloud
- SRequestyRuntimes & gateways · Cloud
- SSGLangRuntimes & gateways · Local
- STogether AIRuntimes & gateways · Cloud
- SVeniceApps · Cloud
- SVercel AI GatewayRuntimes & gateways · Cloud
Sources
- https://huggingface.co/meta-models/Muse-Spark-Open-816B-A42B
- https://huggingface.co/datasets/lmarena-ai/leaderboard-dataset
- https://artificialanalysis.ai/leaderboards/models
- https://openrouter.ai/api/v1/models/meta/muse-spark-open/endpoints
- https://ollama.com/search?c=vision
- https://artificialanalysis.ai/evaluations/mmmu-pro
- https://labs.scale.com/leaderboard/swe_bench_pro_public_v2
- https://artificialanalysis.ai/evaluations/automationbench-aa