Mixtral 8x7B hardware requirements
46.7B parameters (12.9B active, mixture of experts). Needs about 30 GB at Q4_K_M with 8K context. Model card.
Memory needed at each quantization
| Quant | Quality | 8K context | 32K context |
|---|---|---|---|
| FP16 | Full quality | 95.1 GB | 98.3 GB |
| Q8_0 | Near-lossless | 51.3 GB | 54.5 GB |
| Q6_K | Excellent | 40 GB | 43.2 GB |
| Q5_K_M | Very good | 34.9 GB | 38.1 GB |
| Q4_K_M | Good, most popular | 30 GB | 33.2 GB |
| Q3_K_M | Noticeably degraded | 24.5 GB | 27.7 GB |
| Q2_K | Heavily degraded | 19.2 GB | 22.4 GB |
Run it with Ollama: ollama run mixtral:8x7b (downloads Q4_K_M by default).
Run it in the cloud
Don't want to buy hardware? Run Mixtral 8x7B on a rented GPU today.
Rent a cloud GPU by the hour. No upfront cost, cancel any time.
Every GPU and Mac checked
| GPU | Runs? | Best fit |
|---|---|---|
| NVIDIA RTX 3060 12GB | No | Needs ~30 GB at Q4_K_M; 12 GB available |
| NVIDIA RTX 3070 | No | Needs ~30 GB at Q4_K_M; 8 GB available |
| NVIDIA RTX 3080 10GB | No | Needs ~30 GB at Q4_K_M; 10 GB available |
| NVIDIA RTX 3090 | Yes, with heavy quantization | Q2_K · ~19.2 GB · ~116 tok/s |
| NVIDIA RTX 4060 | No | Needs ~30 GB at Q4_K_M; 8 GB available |
| NVIDIA RTX 4060 Ti 16GB | No | Needs ~30 GB at Q4_K_M; 16 GB available |
| NVIDIA RTX 5060 Ti 16GB | No | Needs ~30 GB at Q4_K_M; 16 GB available |
| NVIDIA RTX 4070 | No | Needs ~30 GB at Q4_K_M; 12 GB available |
| NVIDIA RTX 4070 Ti Super | No | Needs ~30 GB at Q4_K_M; 16 GB available |
| NVIDIA RTX 4080 | No | Needs ~30 GB at Q4_K_M; 16 GB available |
| NVIDIA RTX 4090 | Yes, with heavy quantization | Q2_K · ~19.2 GB · ~125 tok/s |
| NVIDIA RTX 5070 | No | Needs ~30 GB at Q4_K_M; 12 GB available |
| NVIDIA RTX 5070 Ti | No | Needs ~30 GB at Q4_K_M; 16 GB available |
| NVIDIA RTX 5080 | No | Needs ~30 GB at Q4_K_M; 16 GB available |
| NVIDIA RTX 5090 | Yes | Q4_K_M · ~30 GB · ~137 tok/s |
| AMD Radeon RX 7800 XT | No | Needs ~30 GB at Q4_K_M; 16 GB available |
| AMD Radeon RX 7900 XTX | Yes, with heavy quantization | Q2_K · ~19.2 GB · ~119 tok/s |
| Apple M4 (16GB) | No | Needs ~30 GB at Q4_K_M; 10.7 GB available |
| Apple M1 (16GB) | No | Needs ~30 GB at Q4_K_M; 10.7 GB available |
| Apple M2 Pro (32GB) | Yes, with heavy quantization | Q2_K · ~19.2 GB · ~25 tok/s |
| Apple M3 Max (64GB) | Yes | Q6_K · ~40 GB · ~23 tok/s |
| Apple M4 Pro (48GB) | Yes | Q5_K_M · ~34.9 GB · ~18 tok/s |
| Apple M4 Max (128GB) | Yes | FP16 · ~95.1 GB · ~13 tok/s |
| Apple M2 Ultra (192GB) | Yes | FP16 · ~95.1 GB · ~19 tok/s |