BenchLeader

MiniMax M2.5 vs Muse Spark Open

Full profiles: MiniMax M2.5 (MiniMax) and Muse Spark Open (Meta): every benchmark, price and rank.

Updated · Data from 14 sources

As of 16 Nov 2026, MiniMax M2.5 and Muse Spark Open are neck and neck on quality (53.7 vs 53.5 on the BenchLeader Index, inside the ±8.1 margin), and MiniMax M2.5 is 3.8× cheaper at $0.525 vs $2.00 per 1M tokens blended. Choose MiniMax M2.5 for coding (51 vs 36); choose Muse Spark Open for fast first answers (9.9 s vs 31.6 s) and knowledge (59 vs 52), reasoning (51 vs 46), and long context (64 vs 61), where it also streams 3×+ faster.

Key facts (as of 16 Nov 2026)

  • MiniMax M2.5 scores 53.7 on the BenchLeader Index and Muse Spark Open scores 53.5 (BenchLeader, 16 Nov 2026).
  • MiniMax M2.5 costs $0.525 per 1M tokens blended (list price, 16 Nov 2026).
  • Muse Spark Open costs $2.00 per 1M tokens blended (list price, 16 Nov 2026).
  • Muse Spark Open streams 3×+ faster than MiniMax M2.5 (Artificial Analysis, 7-day median, 16 Nov 2026).
  • Of the 10 benchmarks both models have, Muse Spark Open leads 9 (BenchLeader, 16 Nov 2026).

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
CodingMiniMax M2.551 vs 36 category score; Vibe Code Bench v1.1 14.8% vs 76.2%High
ReasoningMuse Spark Open51 vs 46 category score; ARC-AGI-2 57.5% vs 4.9%Medium
KnowledgeMuse Spark Open59 vs 52 category scoreMedium
Long documentsMuse Spark Open64 vs 61 category score; AA-LCR v1.1 78.3% vs 73.3%Medium
Chat quality (human preference)–Only MiniMax M2.5 has an Arena rating (56 category score)Not comparable
Lowest costMiniMax M2.5$0.525 vs $2.00 per 1M blended; agentic coding session $0.023 vs $0.092–
Fastest first answerMuse Spark Open9.9 s (Muse Spark Open (xhigh)) vs 31.6 s (MiniMax M2.5) 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

  • MiniMax M2.5 and Muse Spark Open (xhigh) are neck and neck on quality (53.7 vs 53.5, within the ±8.1 margin).
  • MiniMax M2.5 is stronger in coding.
  • Muse Spark Open (xhigh) is stronger in knowledge, long context, reasoning.
  • MiniMax M2.5 is 3.8× cheaper ($0.525 vs $2.00 per 1M blended).
  • Muse Spark Open (xhigh) streams 4.4× faster (180 vs 41 tokens per second).

At a glance

SpecMiniMax M2.5Muse Spark Open
DeveloperMiniMaxMeta
Released12 Feb 202615 Oct 2026
Context window (tokens)204,8001,048,576
Compared atMiniMax M2.5Muse Spark Open (xhigh)
Open weightsYesYes
Price in / out / cached input per 1M$0.300 / $1.20 / $0.030$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.

CategoryMiniMax M2.5Muse Spark Open (xhigh)Leader
Reasoning45.551.3Muse Spark Open
Coding50.736.0MiniMax M2.5
Agents & tools46.6–not comparable
Maths51.8–not comparable
Knowledge51.558.8Muse Spark Open
Instruction following70.3–not comparable
Human preference56.1–not comparable
Multimodal–63.3not comparable
Long context61.263.8Muse Spark Open

On one chart

Each domain on the index's own scale, where 50 is an average model. Click a domain to see the benchmarks behind it. Pick up to 5 models from the list.

2550 avg75ReasoningCodingKnowledgeLong contextMiniMax M2.5, Reasoning: 45.5MiniMax M2.5, Coding: 50.7MiniMax M2.5, Knowledge: 51.5MiniMax M2.5, Long context: 61.2Muse Spark Open, Reasoning: 51.3Muse Spark Open, Coding: 36.0Muse Spark Open, Knowledge: 58.8Muse Spark Open, Long context: 63.8

Not drawn, measured for fewer than two of these models: Agents & tools, Maths, Instruction following, Multimodal, Human preference.

Model, click to add or removeIndex

Where sources disagree

Across the 10 benchmarks both have, MiniMax M2.5 leads 0, Muse Spark Open leads 9 and 1 is a tie.

Overall indexMiniMax M2.5Muse Spark Open (xhigh)Leader
BenchLeader Index53.753.5neck and neck

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. Cost, tokens and time per benchmark task are measured: Artificial Analysis’ averages over its whole Intelligence Index run.

MeasureMiniMax M2.5Muse Spark Open (xhigh)
Blended price per 1M tokens$0.525$2.00
Input / output per 1M$0.300 / $1.20$1.25 / $4.25
Cached input per 1M$0.030$0.150
Price per task: chat reply$0.0005$0.0018
Price per task: summarize a 30-page report$0.0043$0.018
Price per task: code edit$0.0036$0.014
Price per task: agentic coding session$0.023$0.092
Price per task: structured extraction$0.0008$0.0034
Agentic coding session, input mostly cached$0.011$0.043
Cost per benchmark task (AA Index)–$2.02
Output tokens per task (AA Index)–66k
Of which reasoning–32k
Time per task (AA Index)–6 min
Output speed41 tok/s180 tok/s
Time to first answer31.6 s9.9 s
Time to a complete answer, last measured 13 Nov 2026104.7 s39.9 s

Each model is compared at its best-ranked reasoning-effort setting. MiniMax M2.5 scores 53.7 and takes 32 s to its first answer token. Muse Spark Open (xhigh) scores 53.5 and takes 10 s to its first answer token. 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

112 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.

MetricPublisherMiniMax M2.5Muse Spark Open (xhigh)
BenchLeader IndexBenchLeader53.753.5
Agents & tools scoreBenchLeader46.6–
Coding scoreBenchLeader50.736.0
Human preference scoreBenchLeader56.1–
Instruction following scoreBenchLeader70.3–
Knowledge scoreBenchLeader51.558.8
Long context scoreBenchLeader61.263.8
Maths scoreBenchLeader51.8–
Reasoning scoreBenchLeader45.551.3
Multimodal scoreBenchLeader–63.3
Blended price $/MBenchLeader$0.525$2.00
Output speedBenchLeader41 tok/s180 tok/s
Time to first answerBenchLeader31.6 s9.9 s
Context windowBenchLeader204,800 tokens1,048,576 tokens
Reasoning
Arena Hard PromptsArena (LMArena)1418–
LiveBench 2.0 ReasoningLiveBench–73.7%†
GPQA Diamond (AA)Artificial Analysis84.8%89.0%
Humanity's Last Exam (AA)Artificial Analysis20.5%43.9%
HLE-Diamond (AA)Artificial Analysis–26.3%
GPQA Diamond (Vals)Vals AI82.1%–
Kagi LLM BenchmarkKagi LLM Benchmark55.2%–
ARC-AGI-1ARC Prize63.7%89.0%
ARC-AGI-2ARC Prize4.9%57.5%
ARC-AGI-3ARC Prize–0.4%
CritPtArtificial Analysis1.1%14.9%
CritPt v2 hard subset (AA)Artificial Analysis–29.5%
HLE-Diamond (Scale)Scale AI SEAL–28.0%†
Arena ExpertArena (LMArena)1427–
CL-benchEpoch AI Benchmarking Hub11.4%–
CL-bench LifeEpoch AI Benchmarking Hub6.3%–
Coding
Arena CodingArena (LMArena)1447–
Arena WebDevArena (LMArena)13861526†
LiveBench 2.0 CodingLiveBench–55.1%†
SciCode (AA)Artificial Analysis–58.5%
LiveCodeBenchVals AI79.2%–
SWE-bench (Vals)Vals AI74.2%–
SWE-bench Verified (bash only)SWE-bench75.8%†–
SWE-bench Multilingual (bash only)SWE-bench68.3%–
SWE-bench Verified (any scaffold)SWE-bench75.8%†–
Show all 112 benchmarks
MetricPublisherMiniMax M2.5Muse Spark Open (xhigh)
Coding
Code MigrationVals AI–21.2%
Code Migration v1.1Vals AI–26.9%
Vibe Code Bench v1.1Vals AI14.8%76.2%
SWE Atlas v1.0: Codebase QnAScale AI SEAL10.3%–
SWE Atlas v1.0: RefactoringScale AI SEAL19.5%–
SWE Atlas v1.0: Test WritingScale AI SEAL18.6%–
SWE Atlas v1.1Scale AI SEAL–42.3%†
SWE Atlas v1.1: Codebase QnAScale AI SEAL–50.0%†
SWE Atlas v1.1: Test WritingScale AI SEAL–41.1%†
SWE Atlas v1.1: RefactoringScale AI SEAL–35.7%†
SWE-Bench Pro V2Scale AI SEAL–95.0%
SWE-Bench Pro V2 HARDScale AI SEAL–82.3%
ALE-BenchEpoch AI Benchmarking Hub618.2–
Terminal-Bench 4.0 (AA)Artificial Analysis–8.1%
Terminal-Bench 2.1 (AA)Artificial Analysis–76.4%
GameBench 2 OverallDuelLab GameBench 2–55.3†
GameBench 2 by reasoning settingDuelLab GameBench 2–59.3
Design Arena WebsiteDesign Arena12141305
FrontierCode 1.1 (AA)Artificial Analysis–27.6%
FrontierCode 2.0 (AA)Artificial Analysis–7.6%
Agents & tools
Terminal-BenchTerminal-Bench42.7%–
LiveBench Long AgenticLiveBench–12.1%†
LiveBench BrowserLiveBench–26.2%†
Terminal-Bench HardArtificial Analysis34.9%–
τ²-Bench Telecom (AA)Artificial Analysis95.3%–
HiL-BenchScale AI SEAL6.3%59.7%†
GDPval-AA v2.1, normalised scoreArtificial Analysis–45.1%
GDPval-AA v2.1Artificial Analysis–1401
Finance Agent v2Vals AI–61.3%
Finance Agent v2.1Vals AI–60.7%
Harvey's Legal Agent Benchmark v1.1Vals AI–25.0%
Harvey's Legal Agent Benchmark v1Vals AI–16.7%
Legal Research BenchVals AI–38.5%
Tax Agent BenchVals AI–26.3%
Terminal-Bench 2.0 (Vals)Vals AI41.6%–
Terminal-Bench 4.0 (Vals)Vals AI–6.6%
Vending-Bench 2Andon Labs-23.2–
AutomationBenchArtificial Analysis–49.7%
GDP.pdf (AA)Artificial Analysis–18.6%
AA-Briefcase v1.1Artificial Analysis–1280
GDPxGDPx by SemiAnalysis–1.7%†
Terminal-Bench-Science 0.1 (AA)Artificial Analysis–1.9%
Maths
ProofBenchVals AI4.0%–
AIME (Vals)Vals AI88.8%–
ArXivMathMathArena–27.0%
BrokenArXivMathArena–30.1%
Arena MathsArena (LMArena)1392–
Knowledge
LiveBench 2.0 Data AnalysisLiveBench–60.2%†
LiveBench ResearchLiveBench–82.7%†
AA-Omniscience v1.1Artificial Analysis–25.8
MMLU-ProVals AI80.1%–
LegalBenchVals AI80.0%–
CorpFinVals AI59.6%–
TaxEvalVals AI68.2%–
MedQAVals AI92.5%–
PRBench LegalScale AI SEAL–59.8%†
CaseLaw v2Vals AI53.5%–
Excel Modeling BenchmarkVals AI–53.7%
Excel Modeling Benchmark v1.1Vals AI–62.0%
AA-Omniscience v1.1: accuracyArtificial Analysis–42.0%
AA-Omniscience v1.1: non-hallucinationArtificial Analysis–72.0%
Global-MMLU-Lite (AA)Artificial Analysis84.2%–
Instruction following
IFBenchArtificial Analysis71.6%–
MultiChallengeScale AI SEAL–74.2%†
Arena Instruction FollowingArena (LMArena)1381–
Human preference
Arena TextArena (LMArena)1391–
EQ-Bench Creative Writing v3EQ-Bench1362–
Arena Creative WritingArena (LMArena)1356–
Arena Multi-turnArena (LMArena)1397–
Arena Longer QueriesArena (LMArena)1406–
Multimodal
LiveBench DesignLiveBench–58.1%†
MMMU-ProArtificial Analysis–80.9%
SAGEVals AI–85.8%
Humanity's Sixth SenseScale AI SEAL–29.6%†
Long context
AA-LCR v1.1Artificial Analysis73.3%78.3%
MLCRArtificial Analysis–7.8%
Composite
Epoch Capabilities IndexEpoch AI Benchmarking Hub–155.4†
LiveBench 2.0LiveBench–48.5%†
AA Intelligence Index v5.0.1Artificial Analysis–24.9
AA Intelligence Index v4.3.2Artificial Analysis22.838.9
AA Progress Index v1Artificial Analysis266306.7
Vals IndexVals AI–48.6
Vals Finance IndexVals AI–53.0%
Vals Legal IndexVals AI–68.5%
Safety & honesty
FORTRESSScale AI SEAL–7.6%†
DistressBenchScale AI SEAL–87.7%†
Metadata
AA Openness IndexArtificial Analysis27.827.8

Data as of 2026-11-16. Best configuration of each model; every score links to its source on the model pages.

DuelLab GameBench 2, release gb2-20261111-7b598bb13de4 (https://benchmarks.duellab.org/releases/gb2-20261111-7b598bb13de4/), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Scores mapped to BenchLeader model names; not endorsed by DuelLab.

What changed

  • MiniMax M2.5: 1 new result (Global-MMLU-Lite (AA)).
  • Muse Spark Open (xhigh): 1 score revision.
  • Muse Spark Open (xhigh): 1 new result (GameBench 2 by reasoning setting); 1 score revision.
  • Muse Spark Open (xhigh): BenchLeader Index 53.3 → 53.8.
  • Muse Spark Open (xhigh): 1 new result (ARC-AGI-3).
  • Muse Spark Open (xhigh): BenchLeader Index 59.5 → 53.3.
  • MiniMax M2.5: 1 new result (AA Progress Index v1).
  • Muse Spark Open (xhigh): 6 new results (AA Intelligence Index v5.0.1, AA Progress Index v1, HLE-Diamond (AA)…); 3 score revisions; cost per task moved from $0.84/M to $2.02/M (139% up).
  • MiniMax M2.5: BenchLeader Index 53.9 → 53.4.
  • Muse Spark Open (xhigh): 6 new results (MLCR, Code Migration v1.1, Excel Modeling Benchmark v1.1…); cached input price moved from $1.25/M to $0.15/M (88% down).

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

MiniMax M2.5 vs Muse Spark Open: questions

Is MiniMax M2.5 better than Muse Spark Open?
MiniMax M2.5 and Muse Spark Open (xhigh) are neck and neck on quality (53.7 vs 53.5, within the ±8.1 margin). The BenchLeader Index combines every independent quality benchmark; the two are within the margin of error as of 16 Nov 2026, so check the category scores for your use.
MiniMax M2.5 vs Muse Spark Open: which is cheaper, and will that change?
MiniMax M2.5 is cheaper today: $0.525 against $2.00 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
Is MiniMax M2.5 or Muse Spark Open better for coding?
MiniMax M2.5 scores higher in coding (51 vs 36 on the category index, where 50 is average).
Which is faster, MiniMax M2.5 or Muse Spark Open?
Muse Spark Open streams 3×+ faster (180 vs 41 output tokens per second, Artificial Analysis 7-day median).
Which is better for long documents?
Muse Spark Open scores higher on long-context benchmarks (64 vs 61). Muse Spark Open accepts more context: 1,048,576 against 204,800 tokens for MiniMax M2.5.
More questions (3)
Which has lower latency?
Muse Spark Open answers first: 9.9 s against 31.6 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.
Do independent benchmarks agree on MiniMax M2.5 vs Muse Spark Open?
Across the 10 benchmarks both have, MiniMax M2.5 leads 0, Muse Spark Open leads 9 and 1 is a tie.
When was Muse Spark Open released?
Muse Spark Open was released on 15 Oct 2026.