Mistral Small 4 (119B)

Mistral AI · released 2026-03-01 · Apache-2.0 license

Mistral Small 4 (119B/6B-active MoE, Apache-2.0, March 2026) unifies instruct, reasoning, multimodal and coding. Mistral's benchmark figures are published only as charts, so no per-metric scores could be sourced yet.

Key specs

TypeLocal open-weight
Parameters
Architecture
Context window
Knowledge cutoff
Modalitiestext, vision
Recommended backends
Minimum viable rig

Benchmark scores

GPQA Diamond
SWE-bench Verified
AIME
MMLU-Pro
BFCL v3 (tool use)
Composite score
Community ratingNo reviews yet

Strengths & weaknesses

Strengths: Unifies instruct, reasoning, multimodal and agentic coding in one model; Efficient MoE (119B total / 6B active), 256K context, Apache-2.0; Configurable reasoning effort; very low price

Weaknesses: Mistral published benchmarks only as chart images, so no per-metric numbers could be sourced; Vendor claims not yet independently verified