Mathstral Models by MistralAI
Last refreshed 2026-05-19. Next refresh: weekly.
Details
About
Mathstral is a family of large language models (LLMs) created by Mistral AI, focusing on mathematical reasoning and scientific exploration. Built on the Mistral 7B architecture, they inherit its efficient design and are fine-tuned specifically for STEM (Science, Technology, Engineering, and Mathematics) fields 124. These models are adept at handling complex, multi-step logical reasoning tasks and are available under the Apache 2.0 license, facilitating open-source collaboration 148. A prominent characteristic is their enhanced performance with increased inference-time computation, offering scalability for various applications 1. They provide leading reasoning abilities for their size category, as evidenced by industry-standard benchmarks 18. Notably, the Mathstral 7B v0.1 model supports a 32k context window, allowing it to tackle more extensive and intricate mathematical challenges 1. Model weights are accessible via Hugging Face, facilitating research and development efforts 1.
Decision facts
- Best fit
- mathematicsmath-heavy prompts
- Capability starting point
- Mathstral 7B with 32k context
- Lowest tracked input
- Not tracked
- Closest related family
- Ministral
Current Variants
Use-when guidance is based on each model's tracked capabilities, context window, release date, and replacement status.
Use when the workload needs 32k context and 7B parameters.
| Model | Use when | Released | Signals | Status |
|---|---|---|---|---|
| Mathstral 7B | Use when the workload needs 32k context and 7B parameters. | 2024-07 | 32k context7B parameters | Current |
Release Timeline
1 release groupSpecifications(1 models)
| Model | Released | Context | Parameters |
|---|---|---|---|
| Mathstral 7B | 2024-07 | 32k | 7B |






