Devstral 2
Devstral 2 ranks first on SWT-bench.
11
Benchmarks with a score
1
Ranked first, out of five or more models
1
In the top three
Mistral AI Dec 2025 Devstral 2
Open weights Flagship
No Epoch Capabilities Index score yet
- Snorkel Agentic Coding 33.2% #8 /11
- ContextBench 40.4% #4 /9
- SWE-bench Multilingual 61.3% #59 /73
Where it ranks
Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| SWT-bench | 89.1% | #1 of 7 | Official leaderboard | |
| ContextBench | 40.4% | #4 of 9 | Official leaderboard | |
| SWE-bench Verified | 53.8% | #33 of 60 | Official leaderboard | |
| Snorkel Agentic Coding | 33.2% | #8 of 11 | Official leaderboard | |
| SWE-bench Multilingual | 61.3% | #59 of 73 | Lab model card | |
| Code Arena | 1,196 | #123 of 128 | Official leaderboard | |
| IOI (Vals) | 0.67% | #54 of 55 | Official leaderboard |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| PinchBench 2.0 | 69.4% | #33 of 57 | Official leaderboard | |
| PinchBench | 67% | #28 of 48 | Official leaderboard | |
| Pencil Puzzle Bench | 0% | #42 of 53 | Official leaderboard | |
| Terminal-Bench 2.0 | 32.6% | #58 of 63 | Lab model card |
Its lab's number next to a leaderboard
Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.
| Benchmark | Lab says | Board says | Gap |
|---|---|---|---|
| SWE-bench Verified | 72.2% SWE-bench Verified | 53.8% Leaderboard, mini-SWE-agent | +18 pts |
Its launch post
The benchmarks the lab chose to headline, with the scores it gave.