← open-weight

Mistral Small 4

Mistral AI · France
Unified model folding Instruct, reasoning (Magistral) and code (Devstral) into a single 119B MoE under Apache 2.0. 6.5B active params, 256K context, 24 languages, toggleable reasoning effort. Strongest permissive EU option at this scale.
Licence facts
Licence
Apache 2.0
Commercial use
Unrestricted
Derivatives
Allowed
Attribution
Required
Parameters
119B MoE (6.5B active, 128 experts, 4 active)
Context
256K
Multimodal
Text + images (input)
Training data
Not disclosed; 24 languages
Last updated
2026-Q1
Known risks
  • Training data undisclosed
  • Multi-GPU required for bf16 serving at full context
Sources
→ Model card→ Apache 2.0
See also
Mistral API· APIMistral Small· open-weightMistral Large 2· open-weightCodestral 22B· open-weight
Reviewed by Ali Madjaji · Last reviewed 2026-04-15
Home·Methodology·About·/