Ministral 8B

Ministral 8B is an 8-billion-parameter model from Mistral with interleaved sliding-window attention for efficient inference and a 128k token context window.

Strengths

The sliding-window attention pattern reduces memory consumption and inference latency while maintaining a large context window.

Best for

Edge deployments and resource-constrained environments where inference speed and memory efficiency are priorities.

Limitations

At 8B parameters, it trades off reasoning capability and knowledge breadth compared to larger models in Mistral's lineup like Mistral Small 3 (24B) or Mistral Large 3 (41B active).