28.3K Downloads
Description
MistralAI's first reasoning model, based on Mistral Small 3.1
Use cases
Minimum system memory
Tags
Last update
Updated on June 10byREADME
Magistral Small
Magistral Small builds upon Mistral Small 3.1 with added reasoning capabilities through SFT from Magistral Medium traces and RL training. It's a small, efficient reasoning model with 24B parameters that can be deployed locally on a single RTX 4090 or 32GB RAM MacBook once quantized.
The model is capable of long chains of reasoning traces before providing answers and supports dozens of languages including English, French, German, Japanese, Chinese, and many others. It features a 128k context window with an Apache 2.0 license for both commercial and non-commercial use.
Custom Fields
Special features defined by the model author
Enable Thinking
: boolean
(default=true)
Controls whether the model will think before replying
Parameters
Custom configuration options included with this model
Sources
The underlying model files this model uses