Muse Spark Open

Released
2026-10-15
Last refreshed
2026-10-15
Status
Researched today
Open weightsCommercial use: conditionalMultimodalCodingRAGAgentsLong contextVisionJSON / Tool use
AI at Meta releases · 11 in the last 12 months · this family litChangelog →
Specifications
Family
Muse
Released
2026-10-15
Context
1M
Parameters
816B (42B activated)
Architecture
Mixture of Experts
Specialization
coding
Openness
Open weights
License
Muse Model LicenseCommercial use: conditional
Weights
Available
Code
Unknown
Created by

Large-scale open-source AI for social technologies.

Menlo Park, California, United States
Founded 2013
Website
Pricing
Output / 1M
$4.25
Input / 1M
$1.25

Cheapest of 4 routes · Fireworks AI · cache read $0.150

About

Muse Spark Open is the open-weight release of Meta Superintelligence Labs' Muse Spark, an 816-billion-parameter mixture-of-experts model that activates about 42 billion parameters per token. Released October 15, 2026 under the Muse Model License, it is built for coding and agentic work, takes text, image, video and audio input, returns text, and has a 1M-token context window. Weights are on Hugging Face in BF16 and NVFP4.

Providers(4)

Compare all 4

Compare API pricing across 4 tracked provider routes for input and output tokens, batch, and cached reads when available.

ProviderInput / 1MOutput / 1MCacheRoute
Fireworks AI$1.25$4.25read $0.150
Serverless
Together AI$1.25$4.25read $0.150
Serverless
GroqCloudContact salesContact sales-
Serverless
OpenRouter---
ServerlessPartial

Available via routers & gateways(1)

Capabilities

VisionMultimodalReasoningJSON / Tool useAudio

Benchmark peer barsfor Coding

No task-mapped benchmark peers are available for this model yet.

Benchmark scores(9)

Scores are benchmark-specific and are direction-aware: the same numeric gap can mean very different outcomes across suites. Use the leaderboard context and this model's provider route to decide whether the winning margin is meaningful for your workload.
BenchmarkScoreVersionEvaluationSource
Google-Proof Q&A89.0GPQA Diamond; xhigh reasoning effortObserved 2026-10-15—Source
Humanity's Last Exam — With Tools61.0HLE (w/ tools); xhigh reasoning effortObserved 2026-10-15—Source
MCP-Atlas89.2MCP-Atlas; xhigh reasoning effortObserved 2026-10-15—Source
OSWorld-Verified80.4OSWorld Verified; xhigh reasoning effortObserved 2026-10-15—Source
Toolathlon72.8Toolathlon Verified; xhigh reasoning effortObserved 2026-10-15—Source
DeepSWE 1.160.0DeepSWE 1.1, preferred harness; xhigh reasoning effortObserved 2026-10-15—Source
Terminal-Bench 2.176.4Terminal-Bench 2.1, Terminus 2 agent; xhigh reasoning effortObserved 2026-10-15—Source
BabyVision w/ tools80.2BabyVision (w/ tools); xhigh reasoning effortObserved 2026-10-15—Source
WorldVQA49.4WorldVQA (w/ tools); xhighObserved 2026-10-15—Source

Migration checks

No linked migration route is available for this model yet.

API versions

meta-models/Muse-Spark-Open-816B-A42Bmeta-models/Muse-Spark-Open-816B-A42B-NVFP4