Up to four models
MiniMax-M3 vs MiniMax-M2.7
On the index (GPT-4 = 100), MiniMax-M3 scores 147 and MiniMax-M2.7 132. Side by side on the same third-party tests, with price and context. The link in your address bar shares this exact comparison.
147131–164
WeightsOpen
ReleasedJun 2026
Price in / out$0.30 / $1.2
Context1.05M
Core tests taken4 of 9
132114–151
WeightsOpen
ReleasedMar 2026
Price in / out$0.21 / $0.84
Context197K
Core tests taken1 of 9
MiniMax-M3 leads MiniMax-M2.7 by 15 points, ahead on 1 of 1 shared core tests, at 1.4× the price.
Test by testthird-party results
Every result as a table
| Test | MiniMax-M3 | MiniMax-M2.7 |
|---|---|---|
| Reasoning & knowledge | ||
| GPQA Diamond (Vals AI) | 92.7% | 86.6% |
| MMLU Pro | 84.2% | 80.4% |
| Coding | ||
| Code Migration | 19.9% | 8.7% |
| LiveCodeBench | 82.2% | 79.9% |
| SWE-bench | 75.0% | 73.8% |
| Vibe Code Bench v1.1 | 47.6% | 11.9% |
| Professional work | ||
| Finance Agent (v2) | 48.3% | 27.9% |
| CorpFin v2 | 68.1% | 61.2% |
| EMB | 47.8% | 28.4% |
| LegalBench | 85.4% | 84.0% |
| Legal Research Bench | 29.8% | 10.6% |
| MedCode | 46.3% | 34.4% |
| MedScribe | 87.3% | 79.9% |
| Tax Agent Bench | 20.6% | 6.9% |
| TaxEval v2 | 72.7% | 66.6% |
Tests at least two of these models have taken. Bold rows are the core tests.