Skip to main content

Model profile

MiniMax M3

MiniMaxCurrentReleased Jun 1, 2026
Data verified
Overall Score
70Prov. #28 of 78Verified #18 of 32
Arena Elo
1445
Categories Ranked
2of 8
Price (1M tokens)
$0.3 in / $1.2 out
API pricing
Speed
Not listed
Context
1M

Evidence coverage

45 of 313 tracked benchmarks are published. 22 are verified and 23 provisional. 7 of 8 categories are measured.

Updated July 16, 2026Methodology
Published / tracked
45 / 313
Verified
22
Provisional
23
Categories measured
7 / 8

Evidence by category

  • Agentic16 benchmarks
    Mixed evidence
  • Coding11 benchmarks
    Mixed evidence
  • Reasoning2 benchmarks
    Reported
  • Knowledge7 benchmarks
    Reported
  • Math1 benchmark
    Verified
  • Multilingual0 benchmarks
    Not measured
  • Multimodal7 benchmarks
    Mixed evidence
  • Inst. Following1 benchmark
    Reported
Open WeightSelf-hostNon-Reasoning
Confidence:
Medium
base

According to BenchLM.ai, MiniMax M3 ranks #28 out of 78 models on the provisional leaderboard with an overall score of 70/100. It also ranks #18 out of 32 on the verified leaderboard. This places it in the mid-tier of AI models, with strengths in specific benchmark categories.

MiniMax M3 is a open weight model with a 1M token context window. It processes queries without explicit chain-of-thought reasoning, offering faster response times and lower token usage.

BenchLM links it directly to MiniMax M2.7 as the earlier related model in that lineage. This profile currently has 45 of 313 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 Agentic (#20), while its weakest is Multimodal & Grounded (#72). This performance profile makes it particularly useful for coding agents, browser research, and computer-use workflows.

Peer position

Exact provisional scores and ranks for the closest listed peers.

Range 70.071.0

  1. GPT-5.1-Codex-Max
    OpenAI
    #2371.0
    GPT-5.1-Codex-Max is #23 with a score of 71.0.
    Compare
  2. GPT-5.1
    OpenAI
    #2571.0
    GPT-5.1 is #25 with a score of 71.0.
    Compare
  3. Claude Opus 4.5
    Anthropic
    #2671.0
    Claude Opus 4.5 is #26 with a score of 71.0.
    Compare
  4. GPT-5 (high)
    OpenAI
    #2771.0
    GPT-5 (high) is #27 with a score of 71.0.
    Compare
  5. MiniMax M3Current model
    MiniMax
    #2870.0
    MiniMax M3 is #28 with a score of 70.0.
  6. Kimi K2.5 (Reasoning)
    Moonshot AI
    #2970.0
    Kimi K2.5 (Reasoning) is #29 with a score of 70.0.
    Compare
  7. DeepSeek V4 Flash (Max)
    DeepSeek
    #3070.0
    DeepSeek V4 Flash (Max) is #30 with a score of 70.0.
    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. Agentic84%
    Eligible cohort rank #20 of 121Category score 77.9
  2. Multimodal36%
    Eligible cohort rank #72 of 112Category score 49.8

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 #20 of 121Percentile 84thWeight 22%16 benchmarksMixed sources77.9
CodingRank Not rankedWeight 20%11 benchmarksMixed sources72.9
ReasoningRank Not rankedWeight 17%2 benchmarksReported0.0
KnowledgeRank Not rankedWeight 12%7 benchmarksReported0.0
MathRank Not rankedWeight 5%1 benchmarkVerified68.8
MultilingualWeight 7%0 benchmarksNot measuredNot measured
MultimodalRank #72 of 112Percentile 36thWeight 12%7 benchmarksMixed sources49.8
Inst. FollowingRank Not rankedWeight 5%1 benchmarkReported0.0

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 Overall1445±5.523,823
Coding1499±8.36,933
Math1444±18.01,125
Instruction Following1439±7.97,909
Creative Writing1405±10.43,792
Multi-turn1453±9.94,426
Hard Prompts1466±6.515,762
Hard Prompts (English)1476±8.07,646
Longer Query1455±7.510,419

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.

Agentic16 benchmarks
Terminal-Bench 2.0Provider exact
66%Weighted 38%
Source: MiniMax: MiniMax M3Provenance: MiniMax reports MiniMax M3 at 66.0% on Terminal-Bench 2.1.
OSWorld-VerifiedProvider exact
70.1%Weighted 34%
Source: MiniMax: MiniMax M3Provenance: MiniMax reports MiniMax M3 at 70.06% on OSWorld-Verified.
BrowseCompProvider exact
83.5%Weighted 28%
Source: MiniMax: MiniMax M3Provenance: MiniMax reports MiniMax M3 at 83.52% on BrowseComp.
MCP AtlasProvider exact
74.2%Display only
Source: MiniMax: MiniMax M3Provenance: MiniMax reports MiniMax M3 at 74.2% on MCP Atlas.
Claw-EvalProvider exact
74.5%Display only
Source: MiniMax: MiniMax M3Provenance: MiniMax reports MiniMax M3 at 74.5% on Claw-Eval.
AA Agentic IndexReported

Artificial Analysis Agentic Index

35.4%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.
τ²-bench resultsReported

τ²-Bench Tool-Agent-User Evaluation

88.9%Display only
Source: Artificial Analysis: tau2-bench leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
GDPval-AAReported

GDPval-AA normalized

44.7%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
1395Display only
Source: Artificial Analysis: gdpval-aa leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
GDPval rubricsProvider exact
74.7%Display only
Source: MiniMaxAI/MiniMax-M3 model cardProvenance: MiniMax reports MiniMax M3 at 74.7% on GDPval rubrics in the model-card comparison chart. BenchLM stores this separately from AA GDPval Elo and normalized GDPval rows.
BankerToolBenchProvider exact
76.1%Display only
Source: MiniMaxAI/MiniMax-M3 model cardProvenance: MiniMax reports MiniMax M3 at 76.1% on BankerToolBench in the model-card comparison chart.
ResearchClawBenchBenchmark exact
19.8%Display only
Source: ResearchClawBench leaderboardProvenance: ResearchClawBench reports this model as ResearchHarness (MiniMax-M3) in the official Pass@1 leaderboard. BenchLM stores the one-decimal RADS average on the local ResearchClawBench display key and excludes it from weighted rankings.
OSWorld 2.0Benchmark exact
4.6%Display only
Source: OSWorld 2.0 paperProvenance: OSWorld 2.0 reports MiniMax M3 single action on its 500-step main table. BenchLM stores the binary completion score and notes the corresponding partial score was 22.3%.
AA BriefcaseReported

Artificial Analysis Briefcase

1110Display only
Source: Artificial Analysis: aa-briefcase leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
AA EnterpriseOps-GymReported

Artificial Analysis EnterpriseOps-Gym

32.1%Display only
Source: Artificial Analysis: enterprise-ops-gym-aa leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
AA Harvey LABReported

Artificial Analysis Harvey LAB-AA

6.7%Display only
Source: Artificial Analysis: harvey-lab-aa leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
Coding11 benchmarks
SWE-bench VerifiedProvider exact

Software Engineering Benchmark Verified

80.5%Weighted 16%
Source: MiniMax: MiniMax M3Provenance: MiniMax reports MiniMax M3 at 80.5% on SWE-Bench Verified.
SWE-bench ProProvider exact
59%Weighted 10%
Source: MiniMax: MiniMax M3Provenance: MiniMax reports MiniMax M3 at 59.0% on SWE-Bench Pro.
Terminal-Bench 2.0Provider exact
66.0%Display only
Source: MiniMax: MiniMax M3Provenance: MiniMax reports MiniMax M3 at 66.0% on Terminal-Bench 2.1. BenchLM stores this on the existing Terminal-Bench display key in both coding and agentic views.
NL2RepoProvider exact
42.1%Display only
Source: MiniMax: MiniMax M3Provenance: MiniMax reports MiniMax M3 at 42.13 on NL2Repo.
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.
Terminal-Bench HardReported
42.4%Display only
Source: Artificial Analysis: terminalbench-hard leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
AA-SciCodeReported

Artificial Analysis SciCode

45.4%Display only
Source: Artificial Analysis: scicode leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
VIBE V2Provider exact
50.1%Display only
Source: MiniMaxAI/MiniMax-M3 model cardProvenance: MiniMax reports MiniMax M3 at 50.1% on VIBE V2 in the model-card comparison chart.
SVG-BenchProvider exact
63.7%Display only
Source: MiniMaxAI/MiniMax-M3 model cardProvenance: MiniMax reports MiniMax M3 at 63.7% on SVG-Bench in the model-card comparison chart.
KernelBench HardProvider exact
28.8%Display only
Source: MiniMaxAI/MiniMax-M3 model cardProvenance: MiniMax reports MiniMax M3 at 28.8% on KernelBench Hard in the model-card comparison chart.
AA Terminal-Bench 2.1Reported

Artificial Analysis Terminal-Bench v2.1

65.2%Display only
Source: Artificial Analysis: terminalbench-v2-1 leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
Reasoning2 benchmarks
AA-LCRReported

Artificial Analysis Long Context Reasoning

74.0%Display only
Source: Artificial Analysis: artificial-analysis-long-context-reasoning leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
CritPtReported

Critical Physics Tasks

3.7%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.
Knowledge7 benchmarks
Artificial Analysis Intelligence IndexReported
44.4%Display only
Source: Artificial Analysis: artificial-analysis-intelligence-index leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
AA-GPQA DiamondReported

Artificial Analysis GPQA Diamond

92.9%Display only
Source: Artificial Analysis: gpqa-diamond leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
AA-HLEReported

Artificial Analysis Humanity's Last Exam

37.1%Display only
Source: Artificial Analysis: humanitys-last-exam leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
AA-Omniscience IndexReported

Artificial Analysis Omniscience Index

1.4%Display only
Source: Artificial Analysis: omniscience leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
AA-Omniscience AccuracyReported

Artificial Analysis Omniscience Accuracy

15.0%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

16.1%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 Openness IndexReported

Artificial Analysis Openness Index

33.3%Display only
Source: Artificial Analysis: artificial-analysis-openness-index leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
Math1 benchmark
USAMO 2026Provider exact

United States of America Mathematical Olympiad 2026

85.7%Weighted 10%
Source: MiniMax: MiniMax M3Provenance: MiniMax reports MiniMax M3 at 36/42 on USAMO 2026; BenchLM stores the equivalent 85.71% score.
Multimodal7 benchmarks
MMMU-ProProvider exact

Massive Multi-discipline Multimodal Understanding Pro

78.1%Weighted 45%
Source: MiniMax: MiniMax M3Provenance: MiniMax reports MiniMax M3 at 78.1% on MMMU-Pro.
OfficeQA ProProvider exact
45.1%Weighted 30%
Source: MiniMax: MiniMax M3Provenance: MiniMax reports MiniMax M3 at 45.1% on OfficeQA Pro.
OmniDocBench 1.5Provider exact
91.6%Display only
Source: MiniMax: MiniMax M3Provenance: MiniMax reports MiniMax M3 at 91.6% on OmniDocBench.
VideoMMMUProvider exact
84.6%Display only
Source: MiniMax: MiniMax M3Provenance: MiniMax reports MiniMax M3 at 84.6% on Video-MMMU.
Video-MME (with subtitle)Provider exact

Video-MME with subtitle

85.4%Display only
Source: MiniMax: MiniMax M3Provenance: MiniMax reports MiniMax M3 at 85.4% on VideoMME with subtitles.
Design Arena WebsiteReported

Design Arena Website Elo

1294Display only
Source: OpenRouter model benchmarksProvenance: Display-only Design Arena Website Elo synced from OpenRouter model benchmark metadata. It is excluded from BenchLM weighted scoring.
AA-MMMU-ProReported

Artificial Analysis MMMU-Pro

78.6%Display only
Source: Artificial Analysis: mmmu-pro leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
Inst. Following1 benchmark
AA-IFBenchReported

Artificial Analysis IFBench

82.9%Display only
Source: Artificial Analysis: ifbench leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.

MiniMax M3 Family

Base entry

Related Earlier Model

MiniMax M2.7

Frequently Asked Questions

How does MiniMax M3 perform overall in AI benchmarks?

MiniMax M3 currently ranks #28 out of 78 models on BenchLM's provisional leaderboard with an overall score of 70. It also ranks #18 out of 32 on the verified leaderboard. It is created by MiniMax and features a 1M context window.

Is MiniMax M3 good for knowledge and understanding?

MiniMax M3 has visible benchmark coverage in knowledge and understanding, but BenchLM does not currently assign it a global category rank there.

Is MiniMax M3 good for coding and programming?

MiniMax M3 has visible benchmark coverage in coding and programming, but BenchLM does not currently assign it a global category rank there.

Is MiniMax M3 good for mathematics?

MiniMax M3 has visible benchmark coverage in mathematics, but BenchLM does not currently assign it a global category rank there.

Is MiniMax M3 good for reasoning and logic?

MiniMax M3 has visible benchmark coverage in reasoning and logic, but BenchLM does not currently assign it a global category rank there.

Is MiniMax M3 good for agentic tool use and computer tasks?

MiniMax M3 ranks #20 out of 78 models in agentic tool use and computer tasks benchmarks with an average score of 77.9. There are stronger options in this category.

Is MiniMax M3 good for multimodal and grounded tasks?

MiniMax M3 ranks #72 out of 78 models in multimodal and grounded tasks benchmarks with an average score of 49.8. There are stronger options in this category.

Is MiniMax M3 good for instruction following?

MiniMax M3 has visible benchmark coverage in instruction following, but BenchLM does not currently assign it a global category rank there.

Is MiniMax M3 open source?

Yes, MiniMax M3 is an open weight model created by MiniMax, meaning it can be downloaded and run locally or fine-tuned for specific use cases.

Does MiniMax M3 have full benchmark coverage on BenchLM?

Not yet. MiniMax M3 currently has 45 published benchmark scores out of the 313 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 MiniMax M3?

MiniMax M3 has a context window of 1M, which determines how much text it can process in a single interaction.

Last updated: July 16, 2026 · Runtime metrics stay blank until BenchLM has a sourced snapshot.

Don't miss the next GPT moment

Which models moved up, what is new, and what it costs. One email each week.

Free. One email per week.