MiniMax-M3.1-Flash places in the top three on AutomationBench.
MiniMax Oct 2026 Mid-size
Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Humanity's Last Exam | 43.4% | #59 of 154 | Lab model card | |
| LOCA-Bench | 60.3% | #1 of 2 | Lab model card | |
| Extended NYT Connections | 28.3% | #83 of 103 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| DesignArena SVG | 1,282 | #16 of 103 | Official leaderboard | |
| DesignArena 3D Design | 1,318 | #22 of 136 | Official leaderboard | |
| NL2Repo-Bench | 61.2% | #5 of 30 | Lab model card | |
| DesignArena Game Dev | 1,311 | #24 of 143 | Official leaderboard | |
| DesignArena Code | 1,297 | #25 of 143 | Official leaderboard | |
| DesignArena UI Component | 1,293 | #25 of 139 | Official leaderboard | |
| DesignArena Website | 1,278 | #33 of 151 | Official leaderboard | |
| DeepSWE | 73.7% | #13 of 56 | Lab model card | |
| DesignArena Agentic Overall | 1,243 | #19 of 74 | Official leaderboard | |
| Code Arena | 1,585 | #36 of 128 | Official leaderboard | |
| DesignArena React Native | 1,221 | #16 of 56 | Official leaderboard | |
| DesignArena Android (Kotlin) | 1,202 | #14 of 46 | Official leaderboard | |
| Vibe Code Bench | 73.9% | #46 of 120 | Official leaderboard | |
| sudo L7 | 18.9% | #7 of 18 | Official leaderboard | |
| ALE-Bench | 1,495 | #56 of 126 | Official leaderboard | |
| SWE-Bench Pro V2 | 96.6% | #8 of 18 | Official leaderboard | |
| DesignArena Data Visualization | 1,217 | #59 of 118 | Official leaderboard | |
| RSI Arena | 1,002 | #5 of 10 | Official leaderboard | |
| PostTrainBench 1.1 | 39.3% | #1 of 2 | Lab model card | |
| OpenDesign Arena | 81.7 | #11 of 21 | Official leaderboard | |
| SWE-Bench Pro V2 Hard | 84.3% | #10 of 18 | Official leaderboard | |
| DesignArena Video to Website | 1,178 | #12 of 19 | Official leaderboard | |
| Tencent WorkBuddy Bench | 75.6% | #9 of 14 | Official leaderboard | |
| DesignArena Image to HTML | 1,169 | #40 of 51 | Official leaderboard | |
| DAXBench v2 | 61% | #42 of 47 | Official leaderboard | |
| FrontierCS 2 | 8.33% | #18 of 19 | Official leaderboard | |
| SWE Refactor Bench | 23% | #1 of 1 | Lab model card |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| AutomationBench | 62.8% | #2 of 47 | Lab model card | |
| OSWorld 2.0 | 61.7% | #9 of 19 | Lab model card | |
| Agent Arena | -2.70% | #37 of 56 | Official leaderboard | |
| Terminal-Bench 4.0 | 38.1% | #19 of 28 | Lab model card | |
| o11y-bench | 64.1% | #16 of 22 | Official leaderboard | |
| BU Bench V2.1 | 50.6 | #16 of 16 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| FoodTruck Bench | $63,464 | #4 of 59 | Official leaderboard | |
| YC-Bench | $2,389,946 | #6 of 59 | Official leaderboard | |
| Agents' Last Exam | 29.6% | #8 of 31 | Official leaderboard | |
| Tax Agent Bench | 32.1% | #24 of 84 | Official leaderboard | |
| APEX-Accounting | 60% | #20 of 49 | Official leaderboard | |
| APEX-Agents 1.1 | 68.1% | #30 of 65 | Official leaderboard | |
| APEX-1 | 65% | #33 of 71 | Official leaderboard | |
| DesignArena Spreadsheet | 1,153 | #17 of 35 | Official leaderboard | |
| DesignArena PDF | 1,187 | #20 of 37 | Official leaderboard | |
| PRBench | 45.1% | #26 of 48 | Official leaderboard | |
| Legal Research Bench | 34.1% | #52 of 93 | Official leaderboard | |
| Vals Finance Index | 55.9% | #26 of 40 | Official leaderboard | |
| Excel Modeling Benchmark v1.1 | 67.8% | #28 of 42 | Official leaderboard | |
| BankerToolBench | 76.2% | #2 of 3 | Lab model card | |
| GDPx | 4.01% | #14 of 20 | Official leaderboard | |
| Vals Index v2.2 | 53.8% | #26 of 37 | Official leaderboard | |
| Harvey LAB v1.1 (HLAB-Vals-1.1) | 7.50% | #27 of 38 | Official leaderboard | |
| Vals Legal Index | 67.1% | #29 of 38 | Official leaderboard | |
| Vals Finance Agent v2.1 | 58.8% | #31 of 40 | Official leaderboard | |
| Agents' Last Exam v1.1 | 29.5% | #12 of 14 | Official leaderboard | |
| Hebbia Financial Services Benchmark | 47% | #9 of 10 | Official leaderboard | |
| TaxCalcBench v2 | 0% | #32 of 32 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| BabyVision | 91% | #5 of 26 | Lab model card | |
| CharXiv | 82.5% | #22 of 75 | Lab model card | |
| DesignArena Code to Video | 1,199 | #9 of 29 | Official leaderboard | |
| VoxelBench | 1,816 | #22 of 56 | Official leaderboard | |
| Frontier3DBench | 52.5 | #6 of 11 | Official leaderboard | |
| GDP.pdf | 27% | #19 of 19 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| CyberGym | 82.2% | #13 of 34 | Lab model card | |
| nocot-bench (NCRI) | 68.4 | #107 of 185 | Official leaderboard | |
| FORTRESS | 52.1% | #60 of 73 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LiveBench | 63.7% | #19 of 31 | Official leaderboard | |
| SAGE v1.1 (Vals) | 85.3% | #18 of 29 | Official leaderboard | |
| SuperCLUE | 41.4 | #10 of 16 | Official leaderboard |
The Self-Reported Index fits one score to the 10 capability results MiniMax reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 156.4 to 164.8. How we compute it.
Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.
The benchmarks the lab chose to headline, with the scores it gave.