Mistral Small 4
Released 2026-03 · hybrid · 256K tokens · 2 benchmarks · Open weight
Editorial notes
Lanzado 16 marzo 2026 en NVIDIA GTC 2026. MoE 119B totales / 6B activos (128 expertos, 4 activos). Multimodal. Primer Mistral que unifica reasoning + multimodal + agentic coding. Apache 2.0. ~$0.20/M input.
Spec sheet
- Company
- Mistral AI
- Country
- FR
- Type
- hybrid
- Release
- 2026-03
- Context
- 256K tokens
- Params total
- 119B
- Params active (MoE)
- 6B
- License
- apache-2.0
- Quants
- BF16, Q8_0, Q5_K_M, Q4_K_M, Q3_K_M
- Pricing (mistral)
- $0.2/$0.6/M
- Slug
- mistral-small-4
Quick install
3 toolsollama run mistral-small:4-instructBuscar "Mistral-Small-4-Instruct" en LM Studiovllm serve mistralai/Mistral-Small-4-Instruct-2603Note: Q4_K_M ~67 GB VRAM
Pricing comparison
| Provider | Input/MTok | Output/MTok | As of |
|---|---|---|---|
| mistral | $0.20 | $0.60 | 2026-03-16 |
| openrouter | $0.20 | $0.60 | 2026-03-16 |
Benchmarks (2)
Cite this model
BibTeX · APA
BibTeX
@misc{frontier-mistral-small-4,
title = {Mistral Small 4},
author = {{Mistral AI}},
year = {2026},
note = {Frontier Benchmarks AI atlas. Accessed 2026-05-08},
url = {https://frontierbenchmarks.com/models/mistral-small-4}
} APA
Mistral AI (2026). Mistral Small 4 [Large language model]. Frontier Benchmarks AI. Retrieved 2026-05-08, from https://frontierbenchmarks.com/models/mistral-small-4
Citation reflects the atlas page, not the original model paper. For the paper, see the "Resources" section above.