Mistral
Mixtral 8x22B (141B parameters) requires approximately 90.9 GB of VRAM with Q4_K_M quantization. As a Mixture of Experts model with 39B 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 105 GB of VRAM.
Get started
— copy & paste to run locallyCopy-paste commands to run Mixtral 8x22B on your machine.
Run
ollama run mixtral:8x22bQuick 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 | 55.0 GB | Low | — |
Q3_K_S | 3 | 69.1 GB | Low | — |
NVFP4 | 4 | 79.0 GB | Medium | — |
Q4_K_M | 4 | 86.0 GB | Medium | — |
Q5_K_M | 5 | 101.5 GB | High | — |
Q6_K | 6 | 115.6 GB | High | — |
Q8_0 | 8 | 150.9 GB | Very High | — |
F16 | 16 | 289.0 GB | Maximum | — |
Quality benchmarks
Coding
Reasoning
General
Source: official · 2024-04-16
Hardware compatibility
Computing compatibility...
Memory breakdown
Frequently asked questions
Mixtral 8x22B (141B parameters) requires approximately 90.9 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 Mixtral 8x22B with a compatibility score of 63/100. It provides 256 GB of memory and achieves approximately 13.5 tokens per second.
The recommended quantization for Mixtral 8x22B 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 Mixtral 8x22B: NVIDIA H200 141GB (score: 70/100), NVIDIA H200 PCIe 141GB (score: 70/100), AMD Instinct MI300A 128GB (score: 70/100). These provide the best combination of memory, bandwidth, and compute for running this model locally.
Yes, Mixtral 8x22B is well-suited for chat as well as reasoning. It was designed with these use cases in mind.
See also
