MiniMaxMiniMax
MiniMax M3
MiniMax · MiniMax. Released Jun 1, 2026.
MiniMax M3 is a model from MiniMax in the MiniMax family, released Jun 1, 2026. evals.report tracks 10 reported MiniMax M3 benchmark scores across AA-Omniscience: Knowledge and Hallucination Benchmark, SWE-bench Pro, Terminal-Bench 2.1, MCP Atlas, Artificial Analysis Intelligence Index, OSWorld, Design Arena, Vibe Code Bench, and 2 more — each shown with its benchmark, metric, source status, and date, and never combined into a single ranking.
Open10 results
Benchmark results 10
Compare this model| Benchmark | Category | Score | Metric | Status | Date | |
|---|---|---|---|---|---|---|
| AA-Omniscience: Knowledge and Hallucination Benchmark | Reasoning | 1 | AA-Omniscience Index | Official | Jun 1, 2026 | Details |
| SWE-bench Pro | Coding | 59.0% | % resolved | Verified | Jun 1, 2026 | Details |
| Terminal-Bench 2.1 | Agents | 66.0% | task success | Verified | Jun 1, 2026 | Details |
| MCP Atlas | Tool use | 74.2% | pass rate | Verified | Jun 1, 2026 | Details |
| Artificial Analysis Intelligence Index | Reasoning | 55 | Index | Official | Jun 1, 2026 | Details |
| OSWorld | Agents | 70.1% | task success rate | Unverified | Jun 1, 2026 | Details |
| Design Arena | Chat preference | 1321 | Elo | Verified | Jun 1, 2026 | Details |
| Vibe Code Bench | Coding | 47.57% | Overall accuracy | Official | Jun 1, 2026 | Details |
| FrontierCode | Coding | 14.7% | weighted score (Main) | Official | Jun 1, 2026 | Details |
| DeepSWE | Coding | 13.3% | % resolved | Unverified | Jun 1, 2026 | Details |