BenchLeader
Menu
Best LLMsChoose a modelLeaderboardsModelsBenchmarksProvidersReleases12

Muse Spark 1.2 vs Muse Spark 1.3

Full profiles: Muse Spark 1.2 (Meta) and Muse Spark 1.3 (Meta): every benchmark, price and rank.

Updated · Data from 7 sources

As of 24 Oct 2026, Muse Spark 1.2 (xhigh) and Muse Spark 1.3 (max) are neck and neck on quality (63.2 vs 67.1, inside the ±6.4 margin of error) and cost about the same, so speed decides it: Muse Spark 1.2 (xhigh) streams 2.6× faster. Pick Muse Spark 1.2 for maths (67 vs 63); pick Muse Spark 1.3 for agents & tools (66 vs 55).

Muse Spark 1.2 benchmarks xhigh Meta Muse Spark 1.3 benchmarks max MetaOpen in the comparison tool

Which to pick for…

The pick's numbers come first. Category scores on the BenchLeader scale (50 is the average model, 15 points one standard deviation). High confidence means a lead of half a standard deviation or more; under one point is called either.

Use casePickWhyConfidence
CodingMuse Spark 1.365 vs 55 category score; Arena WebDev 1657 vs 1534High
Agents & tool useMuse Spark 1.366 vs 55 category score; τ³-Banking (AA) 50.5% vs 34.9%High
MathsMuse Spark 1.267 vs 63 category score; Arena Maths 1491 vs 1510Medium
ReasoningMuse Spark 1.377 vs 69 category score; CritPt 24.9% vs 17.7%High
KnowledgeMuse Spark 1.373 vs 67 category score; AA-Omniscience 25 vs 27.2Medium
Long documentsMuse Spark 1.367 vs 65 category score; AA-LCR 83.0% vs 79.0%Medium
Images and other mediaEither65 vs 66 category score; Arena Vision 1304 vs 1308Either
Chat quality (human preference)Either69 vs 69 on Arena votesEither
Lowest costEither$2.00 vs $2.00 per 1M blended; agentic coding session $0.092 vs $0.092–
Fastest first answerMuse Spark 1.217.6 s (Muse Spark 1.2 (xhigh)) vs 90.1 s (Muse Spark 1.3 (max)) to the first answer token–
High volume on a budgetNeither flagshipA smaller model is usually the better fit: Cheapest good LLMs, Fastest LLMs–

Key differences

  • Muse Spark 1.2 (xhigh) and Muse Spark 1.3 (max) are neck and neck on quality (63.2 vs 67.1, within the ±6.4 margin).
  • Muse Spark 1.2 (xhigh) is stronger in maths.
  • Muse Spark 1.3 (max) is stronger in agents & tools, coding, composite, human preference, knowledge, long context, multimodal, reasoning.
  • They cost about the same ($2.00 per 1M blended).
  • Muse Spark 1.2 (xhigh) streams 2.6× faster (307 vs 116 tokens per second).

At a glance

SpecMuse Spark 1.2Muse Spark 1.3
DeveloperMetaMeta
Released5 Aug 20262 Sept 2026
Context window (tokens)1,048,5761,048,576
Compared atMuse Spark 1.2 (xhigh)Muse Spark 1.3 (max)
Open weightsNoNo
Price in / out / cached input per 1M$1.25 / $4.25 / $0.150$1.25 / $4.25 / $0.150

Category by category

Category scores on the BenchLeader scale: 50 is the average evaluated model, 15 points is one standard deviation. A gap under one point is called level.

CategoryMuse Spark 1.2 (xhigh)Muse Spark 1.3 (max)Leader
Reasoning69.377.4Muse Spark 1.3
Coding55.264.5Muse Spark 1.3
Agents & tools55.065.6Muse Spark 1.3
Maths67.363.4Muse Spark 1.2
Knowledge66.772.8Muse Spark 1.3
Human preference68.669.1level
Multimodal65.265.7level
Long context64.566.6Muse Spark 1.3
Composite73.883.3Muse Spark 1.3

Where sources disagree

Across the 39 benchmarks both have, Muse Spark 1.2 leads 6, Muse Spark 1.3 leads 33. Every overall index that rates both puts Muse Spark 1.3 ahead.

Overall indexMuse Spark 1.2 (xhigh)Muse Spark 1.3 (max)Leader
BenchLeader Index63.267.1neck and neck
AA Intelligence Index v4.3.239.648.1Muse Spark 1.3
Arena Text14911496Muse Spark 1.3

Each publisher runs its own tasks, harnesses and effort settings, so their overall verdicts can differ; BenchLeader reconciles them on one scale. † measured at another effort setting of the same model (the publisher did not test the one compared here).

Price, speed and latency

List prices blended at three input tokens per output token. Per-task prices use the token counts on the methodology page; suite cost is what Artificial Analysis paid to run its whole Intelligence Index.

MeasureMuse Spark 1.2 (xhigh)Muse Spark 1.3 (max)
Blended price per 1M tokens$2.00$2.00
Input / output per 1M$1.25 / $4.25$1.25 / $4.25
Cached input per 1M$0.150$0.150
Price per task: chat reply$0.0018$0.0018
Price per task: summarize a 30-page report$0.018$0.018
Price per task: code edit$0.014$0.014
Price per task: agentic coding session$0.092$0.092
Price per task: structured extraction$0.0034$0.0034
Cost to run a full benchmark suite$0.975$1.60
Output speed307 tok/s116 tok/s
Time to first answer17.6 s90.1 s
Time to a complete answer19.2 s94.4 s

Per token is not per task: on Artificial Analysis' full suite Muse Spark 1.2 (xhigh) cost $0.975 and Muse Spark 1.3 (max) $1.60, because Muse Spark 1.3 (max) used about 1.6× as many tokens for the same work, which a per-token price hides.

Each model is compared at its best-ranked reasoning-effort setting. Muse Spark 1.2 (xhigh) scores 63.2 and takes 18 s to its first answer token. Muse Spark 1.3 (max) scores 67.1 and takes 90 s to its first answer token; Muse Spark 1.3 (xhigh) scores 64.9 in 30 s, Muse Spark 1.3 scores 59.8 in 3 s. Higher effort means more thinking before the answer: better scores, longer waits and more tokens billed. How reasoning effort changes scores and cost.

Every benchmark

110 benchmarks, grouped by category; each name links to the board and its leaders. † measured at another effort setting of the same model, shown for reference and not counted as a win.

MetricPublisherMuse Spark 1.2 (xhigh)Muse Spark 1.3 (max)
BenchLeader IndexBenchLeader63.267.1
Agents & tools scoreBenchLeader55.065.6
Coding scoreBenchLeader55.264.5
Composite scoreBenchLeader73.883.3
Human preference scoreBenchLeader68.669.1
Knowledge scoreBenchLeader66.772.8
Long context scoreBenchLeader64.566.6
Maths scoreBenchLeader67.363.4
Multimodal scoreBenchLeader65.265.7
Reasoning scoreBenchLeader69.377.4
Blended price $/MBenchLeader$2.00$2.00
Output speedBenchLeader307 tok/s116 tok/s
Time to first answerBenchLeader17.6 s90.1 s
Context windowBenchLeader1,048,576 tokens1,048,576 tokens
Reasoning
SimpleBenchSimpleBench74.5%†–
Arena Hard PromptsArena (LMArena)15121517
LiveBench 2.0 ReasoningLiveBench90.0%†89.7%†
LiveBench 1.0 ReasoningLiveBench90.0%89.7%†
GPQA Diamond (AA)Artificial Analysis90.4%93.5%
Humanity's Last Exam (AA)Artificial Analysis45.5%48.7%
CritPtArtificial Analysis17.7%24.9%
MysteryMechanismVals AI–36.0%
Chess PuzzlesEpoch AI Benchmarking Hub27.0%37.0%
Mystery Game PuzzlesEpoch AI Benchmarking Hub–25.0%
LMCAEpoch AI Benchmarking Hub48.4%53.9%†
DTBenchEpoch AI Benchmarking Hub94.7%96.5%†
BTF-3Epoch AI Benchmarking Hub–14.1%†
Coding
SciCodeSciCode57.4%58.8%
WeirdMLWeirdML60.3%–
Arena CodingArena (LMArena)15411536
Arena WebDevArena (LMArena)15341657
LiveBench 2.0 CodingLiveBench78.2%†90.2%†
LiveBench 1.0 CodingLiveBench77.5%81.1%†
IOIVals AI21.8%56.6%
SWE-bench (Vals)Vals AI86.6%–
Code MigrationVals AI29.9%47.4%
ProgramBenchVals AI0.5%2.5%
SWE-bench MultimodalVals AI–43.1%
Vibe Code Bench 1-100Vals AI–20.5%
Show all 110 benchmarks
MetricPublisherMuse Spark 1.2 (xhigh)Muse Spark 1.3 (max)
Coding
Vibe Code Bench v1.1Vals AI79.1%85.9%
DeepSWE v1.1Epoch AI Benchmarking Hub54.9%–
CursorBenchCursor–41.6%
ALE-BenchEpoch AI Benchmarking Hub–1688.3†
FrontierSWEEpoch AI Benchmarking Hub12.0%25.8%
Terminal-Bench 4.0 (AA)Artificial Analysis7.1%33.3%
Terminal-Bench 2.1 (AA)Artificial Analysis80.2%84.3%
Startrise FrontendStartrise70.3%†74.7%†
FrontierCode 1.1 (AA)Artificial Analysis–40.6%
Agents & tools
Terminal-BenchTerminal-Bench–14.6%†
APEX-AgentsMercor45.4%†47.6%
Arena AgentArena (LMArena)-6.83.4
LiveBench Long AgenticLiveBench21.3%†40.4%†
LiveBench BrowserLiveBench36.0%†80.0%†
LiveBench Agentic CodingLiveBench57.6%64.1%†
Terminal-Bench 2.1 (Vals)Vals AI69.7%79.0%
HiL-BenchScale AI SEAL–66.9%†
τ³-Banking (AA)Artificial Analysis34.9%50.5%
ITBench SRE (AA)Artificial Analysis–33.2%
Analyst Agent (AA)Artificial Analysis–37.5%
CUA-benchVals AI–5.8%
CyberBenchVals AI69.5%72.7%
Finance Agent v2Vals AI60.6%60.0%
Harvey's Legal Agent Benchmark v1Vals AI25.4%23.8%
Legal Research BenchVals AI43.8%55.3%
SkillsBenchVals AI53.0%–
SREBenchVals AI0.0%3.8%
Tax Agent BenchVals AI32.8%42.4%
Terminal-Bench 4.0 (Vals)Vals AI6.1%24.8%
Terminal-Bench ScienceVals AI–10.0%
AutomationBenchArtificial Analysis40.6%57.9%
GDP.pdf (AA)Artificial Analysis17.4%26.6%
Harvey LAB-AA v1.1Harvey–8.9%
EnterpriseOps-GymArtificial Analysis47.3%–
AA-Briefcase v1.1Artificial Analysis–1581
Terminal-Bench-Science 0.1 (AA)Artificial Analysis–10.9%
Maths
FrontierMath Tiers 1–3Epoch AI Benchmarking Hub–74.0%
FrontierMath Tier 4Epoch AI Benchmarking Hub–46.3%
OTIS Mock AIMEEpoch AI Benchmarking Hub–99.2%†
ProofBenchVals AI43.0%†58.0%
LiveBench MathematicsLiveBench91.2%96.0%†
Arena MathsArena (LMArena)14911510
Knowledge
SimpleQA VerifiedEpoch AI Benchmarking Hub60.3%–
LiveBench 2.0 Data AnalysisLiveBench76.5%†79.6%†
LiveBench ResearchLiveBench67.7%†68.6%†
LiveBench 1.0 Data AnalysisLiveBench76.5%79.6%†
AA-OmniscienceArtificial Analysis27.225
MMLU-ProVals AI88.3%–
LegalBenchVals AI85.3%–
CorpFinVals AI70.9%–
TaxEvalVals AI80.4%–
PRBench FinanceScale AI SEAL–59.5%†
PRBench LegalScale AI SEAL–61.6%†
BioMysteryBenchVals AI64.8%–
Excel Modeling BenchmarkVals AI57.0%67.4%
MedCodeVals AI49.4%–
MedScribeVals AI90.1%–
Public Benefits Bench v1.1Vals AI68.5%–
DrugDiscoveryBenchScale AI SEAL–62.2%†
GDP.pdfEpoch AI Benchmarking Hub16.0%27.6%†
AA-Omniscience: accuracyArtificial Analysis45.4%43.6%
AA-Omniscience: non-hallucinationArtificial Analysis66.7%67.1%
Instruction following
LiveBench LanguageLiveBench78.6%82.8%†
LiveBench Instruction FollowingLiveBench74.3%78.0%†
Arena Instruction FollowingArena (LMArena)14721485
Human preference
Arena TextArena (LMArena)14911496
Arena Creative WritingArena (LMArena)14391463
Arena Multi-turnArena (LMArena)15111494
Arena Longer QueriesArena (LMArena)14881497
Multimodal
Arena VisionArena (LMArena)13041308
LiveBench DesignLiveBench67.1%†74.8%†
MMMU-ProArtificial Analysis–82.0%†
MMMU-Pro (Vals)Vals AI86.1%–
MortgageTaxVals AI65.4%–
SAGEVals AI47.7%–
AA-3DArtificial Analysis–1037
Long context
AA-LCRArtificial Analysis79.0%83.0%
Arena DocumentArena (LMArena)–1468
MLCRArtificial Analysis7.2%13.9%
Composite
Epoch Capabilities IndexEpoch AI Benchmarking Hub155.1†157.2†
LiveBench 2.0LiveBench62.4%†74.8%†
LiveBench 1.0LiveBench78.0%81.6%†
AA Intelligence Index v4.3.2Artificial Analysis39.648.1
Vals RSI IndexVals AI–19.6%
Vals Multimodal IndexVals AI69.8%†–

Data as of 2026-10-24. Best configuration of each model; every score links to its source on the model pages.

What changed

  • Muse Spark 1.2 (xhigh): 1 new result (Harvey's Legal Agent Benchmark v1).
  • Muse Spark 1.3 (max): 3 new results (Harvey's Legal Agent Benchmark v1, Vals RSI Index, SWE-bench Multimodal); 1 score revision.
  • Muse Spark 1.2 (xhigh): 2 score revisions.
  • Muse Spark 1.3 (max): 1 score revision.
  • Muse Spark 1.3 (max): 1 score revision.
  • Muse Spark 1.2 (xhigh): BenchLeader Index 62.6 → 63.3.
  • Muse Spark 1.2 (xhigh): 4 new results (AutomationBench, GDP.pdf (AA), MLCR…).
  • Muse Spark 1.2 (xhigh): 3 new results (Chess Puzzles, ProgramBench, SREBench); 1 score revision.
  • Muse Spark 1.3 (max): 2 new results (ProgramBench, SREBench); 1 score revision.
  • Muse Spark 1.2 (xhigh): 3 score revisions.

Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs

Muse Spark 1.2 vs Muse Spark 1.3: questions

Is Muse Spark 1.2 better than Muse Spark 1.3?
Muse Spark 1.2 (xhigh) and Muse Spark 1.3 (max) are neck and neck on quality (63.2 vs 67.1, within the ±6.4 margin). The BenchLeader Index combines every independent quality benchmark; the two are within the margin of error as of 2026-10-24, so check the category scores for your use.
Which is cheaper, Muse Spark 1.2 or Muse Spark 1.3, and will that change?
They cost the same: $2.00 per million tokens blended. No price change has been announced for either.
Is Muse Spark 1.2 better than Muse Spark 1.3 for coding?
Muse Spark 1.3 scores higher in coding (65 vs 55 on the category index, where 50 is average). On Arena WebDev, the top shared coding benchmark, it is 1534 for Muse Spark 1.2 against 1657 for Muse Spark 1.3.
Is Muse Spark 1.2 better than Muse Spark 1.3 for agents and tool use?
Muse Spark 1.3 scores higher in agentic tasks (66 vs 55 on the category index, where 50 is average). On τ³-Banking (AA), the top shared agents & tools benchmark, it is 34.9% for Muse Spark 1.2 against 50.5% for Muse Spark 1.3.
Which is better for long documents?
Muse Spark 1.3 scores higher on long-context benchmarks (67 vs 65). Both accept 1,048,576 tokens of context.
Which has lower latency?
Muse Spark 1.2 (xhigh) answers first: 17.6 s against 90.1 s to the first answer token, as Artificial Analysis measures it. Reasoning effort drives most of this: a higher setting thinks longer before it answers.
Which is faster, Muse Spark 1.2 or Muse Spark 1.3?
Muse Spark 1.2 streams faster: 307 against 116 output tokens per second.
Do independent benchmarks agree on Muse Spark 1.2 vs Muse Spark 1.3?
Across the 39 benchmarks both have, Muse Spark 1.2 leads 6, Muse Spark 1.3 leads 33. Every overall index that rates both puts Muse Spark 1.3 ahead.
When was Muse Spark 1.3 released?
Muse Spark 1.3 was released on 2 Sept 2026.