Mistral AI•Released Oct 10, 2023•Updated May 4, 2025
Description
An extremely efficient 7.3B-parameter open model released by Mistral AI (2023), outperforming older larger models in some benchmarks:contentReference[oaicite:40]{index=40}.
Technical Specifications
Parameters
7
Architecture
Transformer with Grouped-Query Attention (GQA) and Sliding Window Attention (SWA)