Devstral 2 is a state-of-the-art open-source model by Mistral AI specializing in agentic coding. It is a 123B-parameter dense transformer model supporting a 256K context window. Devstral 2 supports exploring...
Tested on 8 benchmarks with 41.2% average. Top scores: LiveBench — Coding (66.8%), LiveBench — Mathematics (52.5%), LiveBench — Language (45.7%).
Regularly refreshed coding problems that avoid data contamination. New problems added monthly to prevent memorization.
LiveBench coding tasks that require multi-step reasoning and tool use. Tests planning and execution of complex coding workflows.
Fresh data analysis tasks testing ability to interpret tables, charts, and statistical data.
Regularly refreshed reasoning problems testing logical deduction, spatial reasoning, and analytical thinking.
Regularly updated math problems that test numerical reasoning, algebra, calculus, and combinatorics.
- Typetext
- Context262K tokens (~131 books)
- ReleasedDec 2025
- LicenseOpen Source
- StatusActive
- Cost / Message~$0.003