Mistral AI's 2025 Model Context Lengths | Generated by AI
As of June 2025, Mistral AI has released several models with varying context lengths. Here’s a summary of some of the key models and their context windows:
- Magistral Small (magistral-small-2506): This open-weight model has a 128k context window, though performance might degrade past 40k, so a recommended maximum length is 40k tokens.
- Magistral Medium (magistral-medium-2506): This API-only reasoning model also has a recommended context window of 40k.
- Mistral Small 3.1 (mistral-small-2503) / Mistral Small 3.2 (mistral-small-2506): These models feature a context window of up to 128,000 tokens.
- Codestral (codestral-2501): This code-focused model boasts a significant context window of 256,000 tokens.
- Mistral Large 2 (mistral-large-2407 / mistral-large-2411): Mistral’s flagship large model has a 128k context window.
- Mistral NeMo (open-mistral-nemo-2407): This 12B model, developed in collaboration with NVIDIA, offers a large context window of up to 128,000 tokens.
- Mistral Medium 3 (mistral-medium-2505): This multimodal model also supports a 128k context window.
It’s important to note that while some models advertise very large context windows (e.g., 128k), the practical performance might be optimized for a slightly smaller range (e.g., 40k for Magistral models).