Inference
EN
Meta

Muse Spark

Muse Spark is a reasoning model from Meta. 29 benchmarks count toward its index, in 7 categories.

  • Closed weights
  • Reasoning
  • Input: text
  • Released 2026-04
RankedEnough results for an index.
IndexAll categories, out of 100.59.8±3.5Median 51.0
CoverageIndex weight that has results.95%
SpeedOutput tokens per second.—Median 67/s
First tokenSeconds to the first output token.—Median 1.3 s
Input / 1MUS dollars per 1M input tokens.—Median $0.60
Output / 1MUS dollars per 1M output tokens.—Median $2.55
ContextMaximum tokens in one request.262KMedian 262K
EloLMArena rating and rank.1474±6 #28Median 1422

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.8 counted55.997 / 218
CodingCode writing and repair.3 counted53.3113 / 209
ReasoningLogic problems and puzzles.2 counted59.867 / 227
MultimodalTasks with images and text.6 counted63.033 / 84
KnowledgeFacts and expert knowledge.5 counted67.831 / 268
MultilingualTasks in many languages.None counted——
InstructionTasks with strict prompt rules.1 counted60.0—
MathMath problems.4 counted60.249 / 168

Results

29 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
AIMEMath96.9% ±0.42 / 95
HealthBench HardKnowledgeAbove fitFar above this model's index.42.8%1 / 13
Terminal-Bench 2.0 · Vals AIAgentic59.6% ±5.27 / 66
Artificial Analysis IFBenchInstruction75.9%16 / 146
Artificial Analysis Omniscience AccuracyKnowledge49.6%21 / 164
FrontierMath-2025-02-28-PrivateMath39.0% ±2.99 / 69
Artificial Analysis MMMU-ProMultimodal80.5%18 / 113
MMMU ProMultimodal87.4% ±0.816 / 92
GPQA diamond · Epoch AIKnowledge89.8% ±2.242 / 217
Artificial Analysis Humanity's Last ExamKnowledge40.7%45 / 222
GPQA Diamond · Vals AIKnowledge89.6% ±1.528 / 135
MMLU ProKnowledge87.3% ±0.329 / 135
FrontierMath-Tier-4-2025-07-01-PrivateMath14.6% ±5.113 / 55
τ²-Bench Tool-Agent-User EvaluationAgentic91.5%38 / 149
Artificial Analysis GPQA DiamondKnowledge88.4%51 / 199
Massive Multi-discipline Multimodal Understanding ProMultimodal80.4%14 / 53
Humanity's Last ExamKnowledge50.4%21 / 77
Critical Physics TasksReasoning11.3%60 / 214
ScreenSpot ProMultimodal84.1%6 / 20
Software Engineering Benchmark VerifiedCoding77.4%35 / 114
OTIS Mock AIME 2024-2025Math88.9% ±4.763 / 200
Artificial Analysis Long Context ReasoningReasoning78.0%70 / 217
CharXiv ReasoningMultimodal86.4%17 / 49
GPQA Diamond · BenchLMKnowledge89.5%33 / 84
ZeroBenchMultimodal33.0%4 / 10
SimpleVQAMultimodal71.3%8 / 18
Claw-EvalAgentic63.8%19 / 42
Humanity's Last Exam without toolsKnowledge42.8%24 / 53
Terminal-Bench 2.0 · BenchLMAgentic59.0%21 / 41
GDPval-AA normalizedAgentic24.9%73 / 133
SWE-benchCoding74.4% ±2.048 / 87
Terminal-Bench HardAgentic45.5%15 / 23
ERQAMultimodal64.7%12 / 18
DeepSearchQAAgentic74.8%17 / 24
Vibe Code Bench v1.1Coding19.7% ±4.092 / 123
SWE-bench ProCoding52.4%64 / 83
CyberGymAgentic43.5%28 / 29
TaxEval v2Finance77.7% ±0.83 / 143
Finance Agent v1.1Finance60.6% ±2.83 / 51
MedCodeHealthcare51.3% ±2.216 / 106
MedScribeHealthcare85.9% ±1.825 / 109
CorpFin v2Finance65.1% ±0.932 / 131
LegalBenchLegal84.2% ±0.437 / 149
CaseLaw v2Legal63.1% ±0.817 / 53
LiveCodeBench ProCoding80.0%—
MedXpertQA MultimodalMultimodal78.4%—
MedXpertQA TextKnowledge52.6%—

Sources

BenchLMCC BY-NC 4.0Vals AINo licence statedEpoch AICC BY

History

5 releases
Index60.6 → 59.8
23 Oct 202612 Nov 2026

Questions

How good is Muse Spark?
Its index is 59.8 out of 100. That is place 76 of 283 ranked models.
What is Muse Spark best at?
Knowledge: 67.8 out of 100, place 31 of 268.
What is the context window of Muse Spark?
262K tokens.
Is Muse Spark open source?
No. Its weights are not published.
When was Muse Spark released?
April 2026.

Badge

Muse Spark on Inference360: place 76 of 283, index 59.8
HTML
<a href="https://inference360.ai/models/muse-spark"><img src="https://inference360.ai/badges/muse-spark.svg" alt="Muse Spark on Inference360: place 76 of 283, index 59.8" height="20"></a>
Markdown
[![Muse Spark on Inference360: place 76 of 283, index 59.8](https://inference360.ai/badges/muse-spark.svg)](https://inference360.ai/models/muse-spark)

Compare with

GPT-5.2-Codex59.9Claude Opus 4.6 (Adaptive)60.0Beam 1.0 501B-A23B FP859.2Qwen3.8-27B59.2Muse Spark 1.164.6

More from Meta

All models
Llama 4 BehemothUnrankedMuse Spark 1.164.6Muse Spark 1.264.9Muse Glimmer 30B53.0Muse Spark 1.2 ContributorUnranked