Mistral
Mistral Small 4 119B (119B parameters) requires approximately 79.5 GB of VRAM with Q4_K_M quantization. As a Mixture of Experts model with 6.5B active parameters, it uses less memory than its total parameter count suggests. For the best balance of quality and speed, we recommend hardware with at least 92 GB of VRAM.
Get started
— copy & paste to run locallyCopy-paste commands to run Mistral Small 4 119B on your machine.
Run
lms load Mistral-Small-4-119B-2603 && lms server startQuick specs
About this model
Related models
Quick picks
Best hardware
Run this model
Quantization options
No hardware detected — fit column shows raw VRAM estimates
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 46.4 GB | Low | — |
Q3_K_S | 3 | 58.3 GB | Low | — |
NVFP4 | 4 | 66.6 GB | Medium | — |
Q4_K_M | 4 | 72.6 GB | Medium | — |
Q5_K_M | 5 | 85.7 GB | High | — |
Q6_K | 6 | 97.6 GB | High | — |
Q8_0 | 8 | 127.3 GB | Very High | — |
F16 | 16 | 244.0 GB | Maximum | — |
Quality benchmarks
Coding
Reasoning
Source: official · 2026-03-16
Hardware compatibility
Computing compatibility...
Memory breakdown
Frequently asked questions
Mistral Small 4 119B (119B parameters) requires approximately 79.5 GB of VRAM with Q4_K_M quantization. Lower quantizations like Q4_K_M use less memory but may reduce quality.
Yes, Mac Studio M3 Ultra 256GB can run Mistral Small 4 119B with a compatibility score of 92/100. It provides 256 GB of memory and achieves approximately 37.6 tokens per second.
The recommended quantization for Mistral Small 4 119B is Q4_K_M, which offers the best balance between model quality and memory efficiency. Higher quantizations preserve more quality but require more VRAM.
The top recommended hardware for Mistral Small 4 119B: AMD Instinct MI250X 128GB (score: 97/100), AMD Instinct MI300A 128GB (score: 97/100), AMD Instinct MI250 128GB (score: 97/100). These provide the best combination of memory, bandwidth, and compute for running this model locally.
Yes, Mistral Small 4 119B is well-suited for chat as well as coding, reasoning, vision. It was designed with these use cases in mind.
See also
