Devstral 2 2512 vs Mistral Medium 3.5
| Benchmark | Devstral 2 2512 | Mistral Medium 3.5 |
|---|---|---|
| Humanity's Last Exam | 3.6 | 13.8 |
| Terminal-Bench 4.0 | 0 | 0 |
| AutomationBench-AA | 3.1 | 6.3 |
| MMMU-Pro (vision) | — | 64.9 |
Specs
| Spec | Devstral 2 2512 | Mistral Medium 3.5 |
|---|---|---|
| Input, per 1M | $0.40 | $1.50 |
| Output, per 1M | $2 | $7.50 |
| Cost per task? | — | $0.43 |
| Speed, tok/s | — | 166 |
| Context | 256K | 262K |
| Weights | Open | Open |
| Released | Dec 2025 | Apr 2026 |
| Memory (4-bit) | 75 GB | 77 GB |
| Licence | custom licence | custom licence |
| Commercial use | Restricted | Restricted |
| See images | — | ✓ |
| Read PDFs | Partial | ✓ |
| Hear audio | — | — |
| Watch video | — | — |
| Call tools | ✓ | ✓ |
| Force a tool call | Partial | Partial |
| Return strict JSON | ✓ | ✓ |
| Search the web | ? | ? |
| Run code | ? | ? |
| Use a computer | ? | ? |
What breaks when you switch
Devstral 2 2512 → Mistral Medium 3.5
Memory (4-bit): 75 GB → 77 GB
Mistral Medium 3.5 → Devstral 2 2512