Mistral AI
Mistral models you can run locally
Mistral AI develops efficient open-weight general, multimodal and coding models, including the Mistral, Ministral and Devstral lines.
Available models
Ordered from the lightest model to the largest.
Ministral 3 3B
3B · Dense
Current-gen tiny Ministral — edge chat with 256K context
256K ctx
Ministral 8B
8B · Dense
Mistral's efficient 8B model
32K ctx
Mistral Nemo 12B
12B · Dense
Multilingual 12B with 128K context
128K ctx
Ministral 3 14B
14B · Dense
Current-gen Ministral mid-size — local assistant with 256K context
256K ctx
Devstral Small 2 24B
24B · Dense
Coding-focused model with 256K context — 68% SWE-bench
256K ctx
Mistral Small 3.1 24B
24B · Dense
Multimodal Mistral with vision support
128K ctx
Mixtral 8x7B
47B · 12.9B active · MoE
MoE with 12.9B active params
32K ctx
Mistral Small 4 119B
119B · 6.5B active · MoE
Sparse Mistral Small 4 — 6.5B active, strong local all-rounder
256K ctx