Autor: Jakub Rusinowski · Ostatnia aktualizacja: 16 marca 2026
Wydanie Mistral AI z 16 marca 2026, scalające dawne linie Magistral/Pixtral/Devstral w jeden model MoE o 119B parametrów łącznie i ~6,5B aktywnych (128 ekspertów, 4 aktywnych na token), z wejściem tekstowym i obrazowym, rozumowaniem oraz agentowym kodowaniem. Licencja Apache 2.0, kontekst 256K. W Q4 mieści się na pojedynczym GPU 24 GB (RTX 4090).
| Licence | What it permits | Applies to |
|---|---|---|
Apache-2.0 | Commercial use permitted Commercial use permitted. No usage restrictions beyond attribution. | Mistral Small 4 119B-A6.5B |
| Mistral Small 4 119B-A6.5B | Min 73 GB VRAM · Q4_K_M · 256,000 ctx · ollama run mistral-small (community GGUF quants; check tag for 119B build) |
The cheapest GPU that runs Mistral Small 4 locally (min 73 GB VRAM) is the AMD Ryzen AI Max+ 395 (96 GB).
Install Ollama then run: ollama run mistral-small (community GGUF quants; check tag for 119B build)
Minimum VRAM: 73 GB. For best results use Q4_K_M quantization.
Mistral Small 4 needs about 73 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: Mistral Small 4 119B-A6.5B (73 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.
Mistral Small 4's smallest variant needs about 73 GB, which exceeds a single RTX 4090 (24 GB). Use multiple GPUs, a higher-VRAM card, or Apple Silicon with large unified memory.
Q4_K_M is the best balance of quality and VRAM for Mistral Small 4 in most cases. Choose Q8_0 for near-lossless quality if you have spare VRAM, or smaller quants (Q3/Q2) only when memory is tight.
Install Ollama, then run: ollama run mistral-small (community GGUF quants; check tag for 119B build). This downloads Mistral Small 4 and starts a local, OpenAI-compatible endpoint — no internet connection is needed after the initial download.