Specifications
Context Window
—
Input Price/1M
—
Output Price/1M
—
Parameters
—
Speed
62 tok/s
Latency (TTFT)
729ms
Benchmarks
Devstral Small 2 results on the main AI model evaluation benchmarks. Higher scores indicate better performance.
Agentic
| Benchmark | Score | Maximum | Methodology |
|---|---|---|---|
| Terminal-Bench Hard | 17.0 | 100.0 | — |
Coding
| Benchmark | Score | Maximum | Methodology |
|---|---|---|---|
| LiveCodeBench | 35.0 | 100.0 | Artificial Analysis official API |
| AA Coding Index | 29.3 | 100.0 | Artificial Analysis official API |
| SciCode | 29.0 | 100.0 | — |
Knowledge
| Benchmark | Score | Maximum | Methodology |
|---|---|---|---|
| MMLU-Pro | 68.0 | 100.0 | — |
Long Context
| Benchmark | Score | Maximum | Methodology |
|---|---|---|---|
| AA-LCR | 24.0 | 100.0 | — |
Math
| Benchmark | Score | Maximum | Methodology |
|---|---|---|---|
| AA Math Index | 34.3 | 100.0 | Artificial Analysis official API |
| AIME 2025 | 34.0 | 100.0 | Artificial Analysis official API |
overall
| Benchmark | Score | Maximum | Methodology |
|---|---|---|---|
| AA Intelligence Index | 17.4 | 100.0 | Artificial Analysis official API |
Reasoning
| Benchmark | Score | Maximum | Methodology |
|---|---|---|---|
| MMLU Pro | 67.8 | 100.0 | Artificial Analysis official API |
| GPQA Diamond | 53.0 | 100.0 | Artificial Analysis official API |
| IFBench | 31.0 | 100.0 | — |
| HLE | 3.0 | 100.0 | — |
Tool Use
| Benchmark | Score | Maximum | Methodology |
|---|---|---|---|
| Tau²-Bench | 23.0 | 100.0 | — |
Information
- Release date
- December 09, 2025
- Tool Calling
- ❌ Not supported
- Vision
- ❌ Not supported
- Audio
- ❌ Not supported
Full Analysis: Devstral Small 2
What is Devstral Small 2?
Devstral Small 2 is an AI model developed by Mistral, classified as a text model. It focuses on text processing and natural language generation. As a proprietary model, it is available via Mistral's cloud API.
Pricing & Costs in 2026
Devstral Small 2 does not have public per-token pricing available at this time. Some models offer access via enterprise plans or research programs. Check Mistral's official website for up-to-date availability and pricing.
Benchmarks & Performance
Devstral Small 2 was evaluated on 14 different benchmarks, covering categories like Agentic, Coding, Knowledge, Long Context, Math, overall, Reasoning, Tool Use. Results show moderate performance across available evaluations.
It's important to note that benchmarks measure specific aspects and don't capture the full user experience. Factors like instruction adherence, behavior in long conversations, and real-world task quality vary significantly between models and aren't always reflected in standard scores.
Recommended Use Cases
Devstral Small 2 specializes in text, offering advanced capabilities for creating and processing text content.
Comparison with Alternatives
In the 2026 AI model ecosystem, Devstral Small 2 competes directly with similarly capable models. Mistral competes in this segment against OpenAI, Anthropic, Google, and Meta. The choice between models depends on the specific use case, budget, latency requirements, and need for features like multimodality and tool calling.
For a detailed side-by-side comparison, use our comparison tool or check the overall model ranking.
Frequently Asked Questions
What is Devstral Small 2?
Devstral Small 2 is an AI model developed by Mistral. It is a text model.
How much does Devstral Small 2 cost?
Devstral Small 2 does not have public per-token pricing available at this time. Check Mistral's official website for up-to-date information.
How does Devstral Small 2 compare with other models?
In available benchmarks, Devstral Small 2 scored: Terminal-Bench Hard: 17/100, LiveCodeBench: 35/100, AA Coding Index: 29.3/100. See the full table above for a detailed comparison.
Is Devstral Small 2 open source?
No, Devstral Small 2 is a proprietary model from Mistral. It is available via cloud API. For open source alternatives, check our open source model ranking.
What is Devstral Small 2 best for?
Devstral Small 2 excels at general-purpose language tasks.
Last updated: July 17, 2026 • View methodology →