Up to four models
MiniMax-M3 vs DeepSeek-V4-Pro (Apr 2026)
On the index (GPT-4 = 100), MiniMax-M3 scores 147 and DeepSeek-V4-Pro (Apr 2026) 147. 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
147131–164
WeightsOpen
ReleasedApr 2026
Price in / out$0.90 / $1.5
Context1.05M
Core tests taken3 of 9
MiniMax-M3 and DeepSeek-V4-Pro (Apr 2026) are too close to call (0 points apart, within the uncertainty), ahead on 0 of 1 shared core tests, and costs 2.0× less.
Test by testthird-party results
Every result as a table
| Test | MiniMax-M3 | DeepSeek-V4-Pro (Apr 2026) |
|---|---|---|
| Reasoning & knowledge | ||
| GPQA Diamond (Epoch AI) | 90.9% | 90.9% |
| LiveBench Data Analysis | 76.2% | 74.5% |
| LiveBench IF | 57.5% | 62.4% |
| LiveBench Language | 76.8% | 78.1% |
| LiveBench Reasoning | 74.5% | 82.7% |
| Maths | ||
| AIME-style maths (OTIS mock) | 71.1% | 96.7% |
| LiveBench Mathematics | 77.0% | 90.7% |
| Chess puzzles | 17.0% | 19.0% |
| Surface Evolver Bench | 55.0% | 40.0% |
| Coding | ||
| LiveBench Agentic Coding | 40.7% | 42.6% |
| LiveBench Coding | 68.2% | 70.0% |
| Agents | ||
| DTBench | 78.9% | 90.7% |
| LMCA | 33.7% | 41.2% |
| Mystery game puzzles | 8.0% | 17.0% |
Tests at least two of these models have taken. Bold rows are the core tests.