# MiniMax M3 Benchmark Scores & Performance

> MiniMax M3 by MiniMax scores 61.51/100 overall, ranking #53 out of 413 AI models.

Canonical page: https://benchlm.ai/models/minimax-m3

Last updated: September 10, 2026

## Model Details

| Property | Value |
|----------|-------|
| Creator | MiniMax |
| Source Type | Open Weight |
| Reasoning Type | Non-Reasoning |
| Context Window | 1M |
| Overall Score | 61.51/100 |
| Overall Rank | #53 of 413 |

## Family & Coverage

- Family: MiniMax M3
- Variant: base
- Benchmarks covered: 47 of 428
- Related earlier model: [MiniMax M2.7](/models/minimax-m2-7)
- Coverage note: BenchLM currently has partial benchmark coverage for this model, so the overall score is conservative.

## Agentic Benchmarks

| Benchmark | Score |
|-----------|-------|
| [Terminal-Bench 2.0](/benchmarks/terminal-bench-2) | 66% |
| [BrowseComp](/benchmarks/browsecomp) | 83.5% |
| [OSWorld-Verified](/benchmarks/osworld-verified) | 70.1% |
| [MCP Atlas](/benchmarks/mcpatlas) | 74.2% |
| [Claw-Eval](/benchmarks/claw-eval) | 74.5% |
| [AA Agentic Index](/benchmarks/aaagenticindex) | 30.8% |
| [τ²-bench results](/benchmarks/tau2-bench) | 88.9% |
| [GDPval-AA](/benchmarks/gdpvalaanormalized) | 40.2% |
| [GDPval-AA](/benchmarks/gdpvalaa) | 1304 |
| [GDPval rubrics](/benchmarks/gdpvalrubrics) | 74.7% |
| [BankerToolBench](/benchmarks/bankertoolbench) | 76.1% |
| [ResearchClawBench](/benchmarks/researchclawbench) | 19.8% |
| [OSWorld 2.0](/benchmarks/osworld2) | 4.6% |
| [AA Harvey LAB](/benchmarks/aaharveylab) | 88.4% |
| [terminalBenchHard](/benchmarks/terminal-bench-hard) | 42.4% |
| [Terminal-Bench 2.1 (Vals)](/benchmarks/valsterminalbench21) | 53.6% |

## Coding Benchmarks

| Benchmark | Score |
|-----------|-------|
| [SWE-bench Verified](/benchmarks/swe-bench-verified) | 80.5% |
| [SWE-bench Pro](/benchmarks/swe-bench-pro) | 59% |
| [Terminal-Bench 2.0](/benchmarks/terminal-bench-2) | 66.0% |
| [NL2Repo](/benchmarks/nl2repo) | 42.1% |
| [AA Coding Index](/benchmarks/aacodingindex) | 58.6% |
| [AA-SciCode](/benchmarks/aascicode) | 47.1% |
| [VIBE V2](/benchmarks/vibev2) | 50.1% |
| [SVG-Bench](/benchmarks/svgbench) | 63.7% |
| [KernelBench Hard](/benchmarks/kernelbenchhard) | 28.8% |
| [OpenHarmony Bench](/benchmarks/openharmonybench) | 48.4% |
| [LiveCodeBench (Vals)](/benchmarks/valslivecodebench) | 82.2% |
| [SWE-bench (Vals)](/benchmarks/valsswebench) | 75.0% |

## Multimodal & Grounded Benchmarks

| Benchmark | Score |
|-----------|-------|
| [OfficeQA Pro](/benchmarks/officeqapro) | 45.1% |
| [OmniDocBench 1.5](/benchmarks/omnidocbench15) | 91.6% |
| [MMMU-Pro](/benchmarks/mmmu-pro) | 78.1% |
| [VideoMMMU](/benchmarks/videommmu) | 84.6% |
| [Video-MME (with subtitle)](/benchmarks/videommewithsub) | 85.4% |
| [Design Arena Website](/benchmarks/designarenawebsite) | 1274 |
| [AA-MMMU-Pro](/benchmarks/aammmupro) | 78.6% |

## Reasoning Benchmarks

| Benchmark | Score |
|-----------|-------|
| [AA-LCR](/benchmarks/lcr) | 83.0% |
| [CritPt](/benchmarks/critpt) | 3.7% |

## Knowledge Benchmarks

| Benchmark | Score |
|-----------|-------|
| [Artificial Analysis Intelligence Index](/benchmarks/artificialanalysis) | 29.6% |
| [AA-GPQA Diamond](/benchmarks/aagpqadiamond) | 92.9% |
| [AA-HLE](/benchmarks/aahle) | 39.0% |
| [AA-Omniscience Index](/benchmarks/aaomniscienceindex) | 1.4% |
| [AA-Omniscience Accuracy](/benchmarks/omniscienceaccuracy) | 16.7% |
| [AA-Omniscience Hallucination Rate](/benchmarks/omnisciencehallucinationrate) | 18.4% |
| [GPQA Diamond (Vals)](/benchmarks/valsgpqadiamond) | 92.7% |
| [MMLU-Pro (Vals)](/benchmarks/valsmmlupro) | 84.2% |

## Instruction Following Benchmarks

| Benchmark | Score |
|-----------|-------|
| [AA-IFBench](/benchmarks/aaifbench) | 82.9% |

## Mathematics Benchmarks

| Benchmark | Score |
|-----------|-------|
| [USAMO 2026](/benchmarks/usamo2026) | 85.7% |

## Other MiniMax Models

- [MiniMax M2.7](/models/minimax-m2-7) - Score: 55.18
- [MiniMax M2.5](/models/minimax-m2-5) - Score: 55.08
- [MiniMax M1 80k](/models/minimax-m1-80k) - Score: 21.56
- [MiniMax Speech-02 HD](/models/minimax-speech-02-hd) - Score: not computed