Venice Atlas
← Back to model catalog

VENICE ATLAS · MODEL DOSSIER · MUSE SPARK OPEN

Model overview

Muse Spark Open

Muse Spark Open is an open-weight reasoning model from Meta for coding and agentic workflows. It supports tool calling and native image understanding in a single checkpoint, using a sparse Mixture-of-Experts architecture with 816B total / 42B active parameters and a 1M-token context window.

muse-spark-open · venice.ai
Text · Venice catalog$1.56 per 1M tokensPrice on Venice AI
ModalityTextOn Venice AI
Primary price$1.56 per 1M tokensper 1M tokens
Context window1.0MMax tokens
Benchmarks17Benchmark results

Benchmark results

See the tests behind each score. They use the benchmark’s original scoring scale.

Agents#3130.56 ratingBiology#1235.59 ratingChemistry#1535.86 ratingCoding#3132.96 ratingVision#2731.23 ratingGeneral intelligence#3244.93 ratingImage To Text#622.49 ratingKnowledge#1010.65 ratingMultimodal#1733.10 ratingPhysics#1535.86 ratingProductivity#116.32 ratingReasoning#3245.23 ratingScience#1210.65 ratingTool Calling#3123.33 rating
Agents2 results
Job Benchjob-bench · agents
61.8%
ReportedOct 2026
Toolathlon-Verifiedtoolathlon-verified · agents
72.8%
ReportedOct 2026
Coding2 results
DeepSWE 1.1deepswe-1.1 · code
60.0%
ReportedOct 2026
SWE Atlas - Codebase QnAswe-atlas-codebase-qna · code
58.0%
ReportedOct 2026
General intelligence2 results
Artificial Analysisartificial-analysis · general
39.0%
ReportedOct 2026
GDPval-AAgdpval-aa · general
46.7%
ReportedOct 2026
Knowledge1 result
AA-Omniscience Indexaa-omniscience-index · knowledge
12.5%
ReportedOct 2026
Multimodal1 result
OSWorld-Verifiedosworld-verified · multimodal
80.4%
ReportedOct 2026
Reasoning5 results
GPQAgpqa · reasoning
89.0%
ReportedOct 2026
Humanity's Last Exam (with tools)humanity's-last-exam-(with-tools) · reasoning
61.0%
ReportedOct 2026
MCP Atlasmcp-atlas · reasoning
89.2%
ReportedOct 2026
SciCodescicode · reasoning
58.5%
ReportedOct 2026
Terminal-Bench 2.1terminal-bench-2.1 · reasoning
76.4%
ReportedOct 2026
Vision4 results
BabyVisionbabyvision · vision
80.2%
ReportedOct 2026
SimpleVQAsimplevqa · vision
73.6%
ReportedOct 2026
WorldVQAworldvqa · vision
49.4%
ReportedOct 2026
ZEROBenchzerobench · vision
50.0%
ReportedOct 2026

Learn what each benchmark measures ↗