Skip to main content

Model profile

Muse Spark

MetaSupersededReleased Apr 8, 2026
Data verified
Superseded:Meta has released newer models in this line —Muse Spark 1.1
Overall Score
71.04Public #13 of 200Verified #11 of 99
Arena Elo
1487
Eligible category ranks
3of 8
Price (1M tokens)
Not listedAPI pricing
Speed
Not listed
Context
262K

Evidence coverage

39 of 323 tracked benchmarks are published. 24 are verified and 15 provisional. 7 of 8 categories are measured.

Updated July 23, 2026Methodology
Published / tracked
39 / 323
Verified
24
Provisional
15
Categories with evidence
7 / 8

Evidence by category

  • Agentic8 benchmarks
    Mixed evidence
  • Coding6 benchmarks
    Mixed evidence
  • Reasoning3 benchmarks
    Mixed evidence
  • Knowledge11 benchmarks
    Mixed evidence
  • Math2 benchmarks
    Verified
  • Multilingual0 benchmarks
    Not measured
  • Multimodal8 benchmarks
    Mixed evidence
  • Inst. Following1 benchmark
    Reported
ProprietaryReasoning
Confidence:
High
base

Muse Spark ranks #13 out of 200 models on the public leaderboard with an overall score of 71.04/100. It also ranks #11 out of 99 on the verified leaderboard. This places it in the upper tier of AI models, with competitive scores across most benchmark categories.

Muse Spark is a proprietary model with a 262K token context window. It uses explicit chain-of-thought reasoning, which typically improves performance on math and complex reasoning tasks at the cost of higher latency and token usage.

Muse Spark sits inside the Muse Spark family alongside Muse Spark 1.1. This profile currently has 39 of 323 tracked benchmarks. BenchLM only exposes non-generated benchmark rows publicly, so missing categories stay blank until a sourced evaluation is available.

Its strongest category is Multimodal & Grounded (#7), while its weakest is Agentic (#14). This performance profile makes it particularly strong for screenshots, documents, charts, and grounded multimodal workflows.

Peer position

Exact provisional scores and ranks for the closest listed peers. A score can appear before a model clears the evidence threshold for a rank, so equal scores can have different rank states.

Range 68.5972.84

  1. Qwen3.7 Max
    Alibaba
    #1072.84
    Qwen3.7 Max is #10 with a score of 72.84.
    Compare
  2. GPT-5.6 Terra
    OpenAI
    #1172.57
    GPT-5.6 Terra is #11 with a score of 72.57.
    Compare
  3. Claude Opus 4.7
    Anthropic
    #1271.94
    Claude Opus 4.7 is #12 with a score of 71.94.
    Compare
  4. Muse SparkCurrent model
    Meta
    #1371.04
    Muse Spark is #13 with a score of 71.04.
  5. MiMo-V2.5-Pro
    Xiaomi
    #1470.19
    MiMo-V2.5-Pro is #14 with a score of 70.19.
    Compare
  6. MiniMax M3
    MiniMax
    #1569.75
    MiniMax M3 is #15 with a score of 69.75.
    Compare
  7. Claude Opus 4.6
    Anthropic
    #1668.59
    Claude Opus 4.6 is #16 with a score of 68.59.
    Compare

Category percentile

More

Relative position among models eligible for each sourced category. A higher percentile means a stronger position within that category's ranked cohort; 100 is highest.

  1. Multimodal79%
    Eligible cohort rank #7 of 29Category score 76.3
  2. Coding90%
    Eligible cohort rank #13 of 122Category score 64.2
  3. Agentic89%
    Eligible cohort rank #14 of 119Category score 59.4

Category evidence

Scores and ranks appear only where this model has published benchmark evidence. Categories without displayable source records remain not measured.

Category scores, ranks, weighting, benchmark coverage, and evidence status
CategoryScore
AgenticRank #14 of 119Percentile 89thWeight 22%8 benchmarksMixed sources59.4
CodingRank #13 of 122Percentile 90thWeight 20%6 benchmarksMixed sources64.2
ReasoningRank Not rankedWeight 17%3 benchmarksMixed sources55.3
KnowledgeRank Not rankedWeight 12%11 benchmarksMixed sources79.0
MathRank Not rankedWeight 5%2 benchmarksVerified56.0
MultilingualWeight 7%0 benchmarksNot measuredNot measured
MultimodalRank #7 of 29Percentile 79thWeight 12%8 benchmarksMixed sources76.3
Inst. FollowingWeight 5%1 benchmarkReportedScore pending

Chatbot Arena performance

Scroll horizontally to inspect confidence intervals and vote counts.

Chatbot Arena Elo, confidence interval, and vote count by evaluation view
ViewEloConfidence intervalVotes
Text Overall1487±5.913,562
Coding1525±10.23,777
Math1461±19.9858
Instruction Following1463±9.34,470
Creative Writing1463±14.11,941
Multi-turn1492±13.52,164
Hard Prompts1504±7.18,707
Hard Prompts (English)1506±9.54,336
Longer Query1473±8.95,259

Benchmark Details

Rows below have a displayable published verification record. Each source link and provenance note remains in the page HTML while its category is closed. Source-unverified manual rows and generated rows stay hidden.

Agentic8 benchmarks
Terminal-Bench 2.0Provider exact
59%Weighted 38%
Source: Meta AI: Muse Spark launch benchmarksProvenance: Provider exact
τ²-bench resultsProvider exact

τ²-Bench Tool-Agent-User Evaluation

91.5%Display only
Source: Meta AI: Muse Spark eval methodologyProvenance: Provider exact
DeepSearchQAProvider exact
74.8%Display only
Source: Meta AI: Muse Spark launch benchmarksProvenance: Provider exact
CyberGymBenchmark exact
43.5%Display only
Source: CyberGym leaderboardProvenance: CyberGym Level 1 reports Meta Agent with Muse Spark on the public leaderboard. BenchLM stores the reported target-vulnerability reproduction success rate on the local cyberGym key.
Claw-EvalBenchmark exact
63.8%Display only
Source: Claw-Eval leaderboardProvenance: Claw-Eval reports this model as muse_spark in the official 2026-05-09 leaderboard snapshot. BenchLM stores the primary Pass^3 value on the local Claw-Eval display key.
AA Agentic IndexReported

Artificial Analysis Agentic Index

28.7%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
GDPval-AAReported

GDPval-AA normalized

32.2%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
GDPval-AAReported
1144Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
Coding6 benchmarks
SWE-bench VerifiedProvider exact

Software Engineering Benchmark Verified

77.4%Weighted 16%
Source: Meta AI: Muse Spark launch benchmarksProvenance: Provider exact
SWE-bench ProProvider exact
52.4%Weighted 10%
Source: Meta AI: Muse Spark launch benchmarksProvenance: Provider exact
LiveCodeBench ProProvider exact
80.0%Display only
Source: Meta AI: Muse Spark launch benchmarksProvenance: Provider exact
Vibe Code BenchBenchmark exact

Vibe Code Bench v1.1

19.67%Display only
Source: Vals AI: Vibe Code Bench v1.1Provenance: Vals Vibe Code Bench v1.1 reports this exact row under meta/muse_spark; BenchLM stores it on the local vibeCodeBench key.
AA Coding IndexReported

Artificial Analysis Coding Index

58.6%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-SciCodeReported

Artificial Analysis SciCode

51.5%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
Reasoning3 benchmarks
ARC-AGI-2Provider exact

Abstraction and Reasoning Corpus for AGI v2

42.5%Weighted 31%
Source: Meta AI: Muse Spark launch benchmarksProvenance: Provider exact
AA-LCRReported

Artificial Analysis Long Context Reasoning

69.7%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
CritPtReported

Critical Physics Tasks

11.3%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
Knowledge11 benchmarks
HLEProvider exact

Humanity's Last Exam

50.4%Weighted 45%
Source: Meta AI: Muse Spark launch benchmarksProvenance: Provider exact
GPQA-DProvider exact

GPQA Diamond

89.5%Display only
Source: Meta AI: Muse Spark launch benchmarksProvenance: Provider exact
HLE w/o toolsProvider exact

Humanity's Last Exam without tools

42.8%Display only
Source: Meta AI: Muse Spark launch benchmarksProvenance: Provider exact
HealthBench HardProvider exact
42.8%Display only
Source: Meta AI: Muse Spark launch benchmarksProvenance: Provider exact
MedXpertQA (Text)Provider exact

MedXpertQA Text

52.6%Display only
Source: Meta AI: Muse Spark launch benchmarksProvenance: Provider exact
Artificial Analysis Intelligence IndexReported
43.1%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-GPQA DiamondReported

Artificial Analysis GPQA Diamond

88.4%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-HLEReported

Artificial Analysis Humanity's Last Exam

39.9%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-Omniscience IndexReported

Artificial Analysis Omniscience Index

4.1%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-Omniscience AccuracyReported

Artificial Analysis Omniscience Accuracy

44.6%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-Omniscience Hallucination RateReported

Artificial Analysis Omniscience Hallucination Rate

73.2%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
Math2 benchmarks
FrontierMath v2 (Tiers 1-3)Benchmark exact

FrontierMath v2 Tiers 1-3

39.000%Weighted 30%
Source: Epoch AI FrontierMath v2 leaderboardProvenance: Epoch AI reports FrontierMath v2 Tiers 1-3 at 39% for muse-spark. BenchLM selects the highest published thinking effort for the model and stores the v2 benchmark slice separately.
FrontierMath v2 (Tier 4)Benchmark exact

FrontierMath v2 Tier 4

14.600%Weighted 10%
Source: Epoch AI FrontierMath v2 leaderboardProvenance: Epoch AI reports FrontierMath v2 Tier 4 at 14.6% for muse-spark. BenchLM selects the highest published thinking effort for the model and stores the v2 benchmark slice separately.
Multimodal8 benchmarks
MMMU-ProProvider exact

Massive Multi-discipline Multimodal Understanding Pro

80.4%Weighted 45%
Source: Meta AI: Muse Spark launch benchmarksProvenance: Provider exact
CharXivProvider exact

CharXiv Reasoning

86.4%Weighted 25%
Source: Meta AI: Muse Spark launch benchmarksProvenance: Provider exact
ERQAProvider exact
64.7%Display only
Source: Meta AI: Muse Spark launch benchmarksProvenance: Provider exact
SimpleVQAProvider exact
71.3%Display only
Source: Meta AI: Muse Spark launch benchmarksProvenance: Provider exact
ScreenSpot ProProvider exact
84.1%Display only
Source: Meta AI: Muse Spark launch benchmarksProvenance: Provider exact
ZeroBenchProvider exact
33.0%Display only
Source: Meta AI: Muse Spark launch benchmarksProvenance: Provider exact
MedXpertQA (MM)Provider exact

MedXpertQA Multimodal

78.4%Display only
Source: Meta AI: Muse Spark launch benchmarksProvenance: Provider exact
AA-MMMU-ProReported

Artificial Analysis MMMU-Pro

80.5%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
Inst. Following1 benchmark
AA-IFBenchReported

Artificial Analysis IFBench

75.9%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.

Muse Spark Family

Base entry

Frequently Asked Questions

How does Muse Spark perform overall in AI benchmarks?

Muse Spark has 39 published benchmark scores on BenchLM, but it does not yet have enough non-generated coverage to receive a global overall rank.

Is Muse Spark good for knowledge and understanding?

Muse Spark has visible benchmark coverage in knowledge and understanding, but BenchLM does not currently assign it a global category rank there.

Is Muse Spark good for coding and programming?

Muse Spark ranks #13 out of 122 models in coding and programming benchmarks with an average score of 64.2. There are stronger options in this category.

Is Muse Spark good for mathematics?

Muse Spark has visible benchmark coverage in mathematics, but BenchLM does not currently assign it a global category rank there.

Is Muse Spark good for reasoning and logic?

Muse Spark has visible benchmark coverage in reasoning and logic, but BenchLM does not currently assign it a global category rank there.

Is Muse Spark good for agentic tool use and computer tasks?

Muse Spark ranks #14 out of 119 models in agentic tool use and computer tasks benchmarks with an average score of 59.4. There are stronger options in this category.

Is Muse Spark good for multimodal and grounded tasks?

Muse Spark ranks #7 out of 29 models in multimodal and grounded tasks benchmarks with an average score of 76.3. It is among the top performers in this category.

Is Muse Spark good for instruction following?

Muse Spark has visible benchmark coverage in instruction following, but BenchLM does not currently assign it a global category rank there.

Which sibling models are related to Muse Spark?

Muse Spark belongs to the Muse Spark family. Related variants on BenchLM include Muse Spark 1.1.

Does Muse Spark have full benchmark coverage on BenchLM?

Not yet. Muse Spark currently has 39 published benchmark scores out of the 323 benchmarks BenchLM tracks. BenchLM only exposes non-generated public benchmark rows, so missing categories stay blank until a sourced evaluation is available.

What is the context window size of Muse Spark?

Muse Spark has a published context window of 262K, which determines how much text it can process in a single interaction.

Last updated: July 23, 2026 · Runtime metrics stay blank until BenchLM has a sourced snapshot.

Choose with this week’s evidence

Join 2,000+ readers for ranking moves, new releases, pricing changes, and the evidence behind them.

Free. One email per week.