Configure a rig →
Mistral47B params13B active (MoE)33k contextpopular

Mixtral 8x7B locally

Mixtral 8x7B is an open-weight LLM from the Mistral family with 47B parameters. Main use: chat, RAG and general assistance. Detected minimum hardware: RTX 5090 (32 GB).

Check on my GPU →Hugging Face ↗
Technical facts
Parameters47B
Max context33k
Q4_K_M29.5 GB
Q5_K_M36.1 GB
Q852.5 GB
FP16105.1 GB
FamilyMistral
Last sync2026-05-12

Available quantizations

Q4_K_M
29.5GB

Acceptable. Good compromise when VRAM is limited.

Q5_K_M
36.1GB

Good quality. Sweet spot for size and precision.

Q8
52.5GB

Near-FP16 quality. Comfortable for production.

FP16
105.1GB

Reference precision. Maximum quality, doubled VRAM.

Compatible GPUs

GPUs that can run Mixtral 8x7B on a single card, ranked by VRAM headroom.

RTX 5090
NVIDIA
32 GB · RTX 50
29.5 / 32 GBtight · Q4
RTX 5000 Ada
NVIDIA
32 GB · RTX Ada
29.5 / 32 GBtight · Q4
Radeon Pro W6800
AMD
32 GB · Radeon Pro W
29.5 / 32 GBtight · Q4
Radeon Pro W7800
AMD
32 GB · Radeon Pro W
29.5 / 32 GBtight · Q4
Instinct MI100
AMD
32 GB · Instinct CDNA 1-2
29.5 / 32 GBtight · Q4
Mac Mini M2 Pro (32GB)
Apple
32 GB · Mac Mini
29.5 / 32 GBtight · Q4
Mac Mini M4 (32GB)
Apple
32 GB · Mac Mini
29.5 / 32 GBtight · Q4
MacBook Air M4 (32GB)
Apple
32 GB · MacBook Air
29.5 / 32 GBtight · Q4
MacBook Pro 14 M1 Pro (32GB)
Apple
32 GB · MacBook Pro 14
29.5 / 32 GBtight · Q4
MacBook Pro 14 M1 Max (32GB)
Apple
32 GB · MacBook Pro 14
29.5 / 32 GBtight · Q4
MacBook Pro 14 M2 Pro (32GB)
Apple
32 GB · MacBook Pro 14
29.5 / 32 GBtight · Q4
MacBook Pro 14 M2 Max (32GB)
Apple
32 GB · MacBook Pro 14
29.5 / 32 GBtight · Q4

Recommended multi-GPU rigs

For Mixtral 8x7B at higher quantization or with more context, a multi-GPU rig gives more headroom.

4× GTX 1070
NVIDIA
32 GB · GTX 10
29.5 / 32 GBtight · Q4
4× GTX 1070 Ti
NVIDIA
32 GB · GTX 10
29.5 / 32 GBtight · Q4
4× GTX 1080
NVIDIA
32 GB · GTX 10
29.5 / 32 GBtight · Q4
4× RTX 2060 Super
NVIDIA
32 GB · RTX 20
29.5 / 32 GBtight · Q4
4× RTX 2070
NVIDIA
32 GB · RTX 20
29.5 / 32 GBtight · Q4
4× RTX 2070 Super
NVIDIA
32 GB · RTX 20
29.5 / 32 GBtight · Q4
4× RTX 2080
NVIDIA
32 GB · RTX 20
29.5 / 32 GBtight · Q4
4× RTX 2080 Super
NVIDIA
32 GB · RTX 20
29.5 / 32 GBtight · Q4

Recommended rig

4× GTX 1070

Mixtral 8x7B with Ubuntu, vLLM, Open WebUI and the model already downloaded.

Configure →

Similar models

VRAM estimates: parameters x bits/8 plus margin. Real performance varies by engine, context length and batch size.
sync: 2026-05-12