BlogLLM Benchmark Scores as Published, Side by Side
MiniMax M3: Benchmark Scores as Published
Every result the catalog holds for MiniMax M3: 49 tests in 11 areas, from 5 pages. 32 printed by MiniMax itself, the rest by the company named beside each number. Each number is as printed; nothing is ranked or rescored.
Developer MiniMaxReleased 2026-06-01Context window 1MInput Text, image, videoOutput TextWeights OpenData version 2026103001, as of 2026-10-30
Knowledge Work
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| AutomationBench | 20.5% | IFM, IFM K2-Horizon-375B-A23B model cardchecked 2026-10-15 | - |
| APEX-Agents | 27.7% | MiniMax, MiniMax M3checked 2026-10-15; printed on 1 more page | - |
| GDPval-AA (Elo) | 1380 | IFM, IFM K2-Horizon-375B-A23B model cardchecked 2026-10-15 | - |
| OfficeQA Pro | 45.1% | MiniMax, MiniMax M3checked 2026-10-15 | Harness: Claude Code |
| GDPval rubrics (pointwise rubric score) | 74.78% | MiniMax, MiniMax M3checked 2026-10-15 | - |
| SpreadsheetBench v1 | 89.35% | MiniMax, MiniMax M3checked 2026-10-15 | Harness: Claude Code |
| YC-Bench (final assets, USD) | $2100000 | MiniMax, MiniMax M3checked 2026-10-15 | - |
Agentic Coding
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| SWE-bench Verified | 80.5% | MiniMax, MiniMax M3checked 2026-10-15 | Harness: Claude Code |
| SWE-Bench Pro | 59% | MiniMax, MiniMax M3checked 2026-10-15; printed on 2 more pages | Harness: Claude Code |
| Terminal-Bench 2.1 | 66% | MiniMax, MiniMax M3checked 2026-10-15; printed on 2 more pages | Harness: Terminus 2 |
| DeepSWE v1.1 | 20% | Z.ai, GLM-5.2 model cardchecked 2026-10-15 | Harness: mini-swe-agent; Context: 400000 |
| NL2Repo-Bench | 42.13 | MiniMax, MiniMax M3checked 2026-10-15; printed on 1 more page | - |
| SWE-Atlas-QnA (setting not stated) | 37.9% | MiniMax, MiniMax M3checked 2026-10-15 | - |
| SWE-Atlas-Test Writing | 30.83% | MiniMax, MiniMax M3checked 2026-10-15 | - |
| SWE-fficiency | 34.8% | MiniMax, MiniMax M3checked 2026-10-15 | Harness: Claude Code |
| VIBE-V2 (MiniMax internal) | 50.12% | MiniMax, MiniMax M3checked 2026-10-15 | Harness: Claude Code |
| SWE-Atlas-QnA (strict) | 42.3% | IFM, IFM K2-Horizon-375B-A23B model cardchecked 2026-10-15 | Tools: no internet (strict) |
ML Engineering
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| PaperBench | 52.6 | MiniMax, MiniMax M3checked 2026-10-15 | Harness: Claude Code |
| KernelBench Hard (TFLOPs relative to hardware peak) | 28.8% | MiniMax, MiniMax M3checked 2026-10-15 | Harness: Claude Code |
| PostTrainBench (version not stated) | 37.1% | MiniMax, MiniMax M3checked 2026-10-15 | Harness: Claude Code |
Science and Math
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| GPQA Diamond | 92.9% | IFM, IFM K2-Horizon-375B-A23B model cardchecked 2026-10-15; printed on 1 more page | - |
| SciCode | 45.4% | IFM, IFM K2-Horizon-375B-A23B model cardchecked 2026-10-15 | - |
| HMMT November 2025 | 84.4% | Z.ai, GLM-5.2 model cardchecked 2026-10-15 | - |
| HMMT February 2026 | 84.4% | Z.ai, GLM-5.2 model cardchecked 2026-10-15 | - |
| CritPt | 3.7% | IFM, IFM K2-Horizon-375B-A23B model cardchecked 2026-10-15; printed on 1 more page | - |
| IMO 2025 (points out of 42) | 35 | MiniMax, MiniMax M3checked 2026-10-15 | - |
| USAMO 2026 (points out of 42) | 36 | MiniMax, MiniMax M3checked 2026-10-15 | - |
Long Context
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| AA-LCR | 80.3% | IFM, IFM K2-Horizon-375B-A23B model cardchecked 2026-10-15 | - |
| LOCA-Bench (256k environment description) | 49.3% | MiniMax, MiniMax M3checked 2026-10-15 | - |
Computer Use
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| OSWorld-Verified | 70.06% | MiniMax, MiniMax M3checked 2026-10-15 | - |
Multimodal Understanding
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| MMMU-Pro | 78.1% | MiniMax, MiniMax M3checked 2026-10-15 | - |
| OmniDocBench 1.5 | 91.6% | MiniMax, MiniMax M3checked 2026-10-15 | - |
| Video-MMMU | 84.6% | MiniMax, MiniMax M3checked 2026-10-15 | - |
| Video-MME (with subtitles) | 85.4% | MiniMax, MiniMax M3checked 2026-10-15 | - |
General Knowledge and Reasoning
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Humanity's Last Exam (no tools) | 39% | IFM, IFM K2-Horizon-375B-A23B model cardchecked 2026-10-15; printed on 1 more page | Tools: none |
| AA-Omniscience Accuracy | 17% | IFM, IFM K2-Horizon-375B-A23B model cardchecked 2026-10-15 | - |
| AA-Omniscience Non-Hallucination | 82% | IFM, IFM K2-Horizon-375B-A23B model cardchecked 2026-10-15 | - |
| CL-bench | 20.48% | MiniMax, MiniMax M3checked 2026-10-15 | - |
Code Generation
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| LiveSQLBench (Base-Full v1) | 40.17% | MiniMax, MiniMax M3checked 2026-10-15 | Harness: Claude Code |
| SVG-Bench (MiniMax internal) | 63.7% | MiniMax, MiniMax M3checked 2026-10-15 | Harness: Claude Code |
Tool Use and Workflows
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| τ³-bench Banking | 15.3% | IFM, IFM K2-Horizon-375B-A23B model cardchecked 2026-10-15 | - |
| Toolathlon Verified | 53.7% | IFM, IFM K2-Horizon-375B-A23B model cardchecked 2026-10-15 | - |
| MCP Atlas (public set) | 74.2% | MiniMax, MiniMax M3checked 2026-10-15; printed on 1 more page | - |
| MCPMark | 48.8% | IFM, IFM K2-Horizon-375B-A23B model cardchecked 2026-10-15 | - |
| Claw-Eval pass^3 | 74.5% | MiniMax, MiniMax M3checked 2026-10-15 | - |
| BankerToolBench | 76.12% | MiniMax, MiniMax M3checked 2026-10-15 | - |
| WildClawBench | 56.4% | IFM, IFM K2-Horizon-375B-A23B model cardchecked 2026-10-15 | - |
Web Search and Research
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| BrowseComp | 83.52% | MiniMax, MiniMax M3checked 2026-10-15; printed on 1 more page | Tools: web browsing |
| DRACO | 73.23% | MiniMax, MiniMax M3checked 2026-10-15 | - |
Price
API list prices in US dollars per one million tokens, as printed on the developer's pricing page on the day checked. Cost per task is measured by Vals AI.
| Price | Value | Printed by | Setup as printed |
|---|---|---|---|
| Input | $0.3 | MiniMax, MiniMax API pricingchecked 2026-10-19 | Context: up to 512,000 tokens; Long context: $0.6 per 1M tokens above 512,000 tokens |
| Cached input | $0.06 | MiniMax, MiniMax API pricingchecked 2026-10-19 | Context: up to 512,000 tokens; Long context: $0.12 per 1M tokens above 512,000 tokens |
| Output | $1.2 | MiniMax, MiniMax API pricingchecked 2026-10-19 | Context: up to 512,000 tokens; Long context: $2.4 per 1M tokens above 512,000 tokens |
| Input, priority | $0.45 | MiniMax, MiniMax API pricingchecked 2026-10-19 | Context: up to 512,000 tokens; Long context: $0.9 per 1M tokens above 512,000 tokens |
| Cached input, priority | $0.09 | MiniMax, MiniMax API pricingchecked 2026-10-19 | Context: up to 512,000 tokens; Long context: $0.18 per 1M tokens above 512,000 tokens |
| Output, priority | $1.8 | MiniMax, MiniMax API pricingchecked 2026-10-19 | Context: up to 512,000 tokens; Long context: $3.6 per 1M tokens above 512,000 tokens |
| Cost per taskVals AI, USD, lower is better | $6.45 | Vals AI, Vibe Code Bench v1.1 leaderboardchecked 2026-10-27 | Harness: OpenHands |
More from MiniMax
Every number here is read from the same catalog file as the benchmark table, data version 2026103001. Where one page printed the model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.