In short
How they differ.
Worked out from the figures below. Prices and specs are the makers' own, read Oct 20, 2026.
- MiniMax-M3 is cheaper per output token: $1.20 against $7.50 per million tokens, so Mistral Medium 3.5 costs 6.3 times as much.
- MiniMax-M3 reads more at once: a context window of 1M tokens against 256K.
- Both can be downloaded and run on your own machine.
- No independent board we follow ranks both models yet.
Side by side
The specs.
| MiniMax-M3 | Mistral Medium 3.5 | |
|---|---|---|
| Made by | MiniMax · China | Mistral AI · France |
| Released | Not stated | Apr 28, 2026 |
| Price per 1M tokens | $0.30 in / $1.20 out | $1.50 in / $7.50 out |
| Context window | 1M | 256K |
| Takes in | text | text, image |
| Download and run it yourself | Yes (minimax-community (license tag shown on Hugging Face)) | Yes (Modified MIT) |
| Size | ~428B parameters, ~23B active | 128B (dense) parameters |
| Maker says | “MiniMax-M3 is a native multimodal model with 1M context.” | “Our frontier-class multimodal model optimized for agentic and coding use cases.” |
| Sources | Prices, Oct 20, 2026 · Description | Prices, Oct 20, 2026 · Description |
Independent tests
Where they place.
Position on each board we follow (1 is best). Epoch AI's and Vectara's scores are shown; the other boards are linked for their own figures.
| Board | MiniMax-M3 | Mistral Medium 3.5 |
|---|---|---|
| GPQA Diamond (run by Vals AI, archived)Vals AI, Sep 1, 2026 | #15 | Not listed |
More comparisons