Mistral · 32B to 80B · Mixture of experts
Mixtral 8x7B Instruct V0.1
46.7 billion parameters in total, but only 12.88 billion are read for each token. That split is the whole point of the design: it costs the memory of a large model and the speed of a small one. It uses grouped-query attention with 8 KV heads for 32 query heads, which cuts the cache to 25% of what full attention would need.
About Mixtral 8x7B Instruct V0.1
The mixture-of-experts model that popularized the design in December 2023: 8 experts, 2 used for each token, Apache 2.0. It needs the memory of a 47B model and reads about 13B of it per token.
Mixtral 8x7B Instruct V0.1 is a mixture-of-experts model with 46.7B parameters, 12.88B of them active for each token, built from 32 layers with a hidden size of 4096. Its repository was published on 2023-12-10. The license listed on Hugging Face is apache-2.0. It uses grouped-query attention with 8 KV heads for 32 query heads, which cuts the cache to 25% of what full attention would need.
Memory needed, by quantization
At an 8k context window. Quality is our rough ranking of how much the compression costs you: anything at or above Q5 is hard to tell apart from the original in normal use.
| Quantization | File size | KV cache | Total VRAM | Quality | What it costs you |
|---|---|---|---|---|---|
| Q8_0 | 46.21 GB | 1 GB | 47.49 GB | 99% | Lossless in practice. Use it when the memory is there. |
| Q6_K | 35.66 GB | 1 GB | 36.97 GB | 98% | Very close to Q8 for two thirds of the size. |
| Q5_K_M | 30.93 GB | 1 GB | 32.25 GB | 96% | The quality-first choice when Q6 will not fit. |
| Q4_K_M | 26.26 GB | 1 GB | 27.59 GB | 93% | The default. Best size-to-quality ratio for local use. |
| IQ4_XS | 23.11 GB | 1 GB | 24.45 GB | 91% | Importance-matrix 4-bit. Q4_K_S quality, smaller file. |
| Q3_K_M | 21.26 GB | 1 GB | 22.6 GB | 86% | Degradation starts to show. A way to fit one size up. |
| IQ3_XXS | 16.64 GB | 1 GB | 17.99 GB | 79% | Aggressive. Only worth it on very large models. |
What a longer conversation costs
Same model at Q4_K_M, only the context window changes. The cache grows in a straight line with every token in the window.
| Context | KV cache | Total VRAM | Fits in 8 GB | Fits in 12 GB | Fits in 24 GB |
|---|---|---|---|---|---|
| 4k | 0.5 GB | 27.09 GB | no | no | no |
| 8k | 1 GB | 27.59 GB | no | no | no |
| 16k | 2 GB | 28.59 GB | no | no | no |
| 32k | 4 GB | 30.59 GB | no | no | no |
The "fits" columns allow for the roughly 0.8 GB Windows keeps for the desktop.
Which hardware runs Mixtral 8x7B Instruct V0.1
33 of 118 consumer devices run it at a quantization worth using, at an 8k context window. Another 3 can load it only by compressing the weights far enough to damage the model, marked with a warning below.
| Device | Memory | Verdict | Quantization | Used | Speed |
|---|---|---|---|---|---|
| RTX PRO 6000 Blackwell NVIDIA | 96 GB | Runs great | Q8_0 | 47.49 GB | 55.8 t/s |
| Apple M3 Ultra 96GB Apple | 96 GB | Runs well | Q8_0 | 47.59 GB | 23.6 t/s |
| Apple M3 Ultra 256GB Apple | 256 GB | Runs well | Q8_0 | 47.59 GB | 23.6 t/s |
| Apple M3 Ultra 512GB Apple | 512 GB | Runs well | Q8_0 | 47.59 GB | 23.6 t/s |
| Apple M1 Ultra 64GB Apple | 64 GB | Just fits | Q8_0 | 47.59 GB | 23 t/s |
| Apple M1 Ultra 128GB Apple | 128 GB | Runs well | Q8_0 | 47.59 GB | 23 t/s |
| Apple M2 Ultra 64GB Apple | 64 GB | Just fits | Q8_0 | 47.59 GB | 23 t/s |
| Apple M2 Ultra 128GB Apple | 128 GB | Runs well | Q8_0 | 47.59 GB | 23 t/s |
| Apple M2 Ultra 192GB Apple | 192 GB | Runs well | Q8_0 | 47.59 GB | 23 t/s |
| RTX 6000 Ada Generation NVIDIA | 48 GB | Runs great | Q6_K | 36.97 GB | 37.9 t/s |
| RTX A6000 NVIDIA | 48 GB | Runs great | Q6_K | 36.97 GB | 30.3 t/s |
| Radeon PRO W7900 AMD | 48 GB | Runs great | Q6_K | 36.97 GB | 28.9 t/s |
| Apple M4 Max 64GB Apple | 64 GB | Just fits | Q8_0 | 47.59 GB | 15.7 t/s |
| Apple M4 Max 128GB Apple | 128 GB | Runs well | Q8_0 | 47.59 GB | 15.7 t/s |
| GeForce RTX 5090 NVIDIA | 32 GB | Runs great | Q4_K_M | 27.59 GB | 93 t/s |
| Apple M1 Max 64GB Apple | 64 GB | Just fits | Q8_0 | 47.59 GB | 11.5 t/s |
| Apple M2 Max 64GB Apple | 64 GB | Just fits | Q8_0 | 47.59 GB | 11.5 t/s |
| Apple M2 Max 96GB Apple | 96 GB | Runs well | Q8_0 | 47.59 GB | 11.5 t/s |
| Apple M3 Max 64GB Apple | 64 GB | Just fits | Q8_0 | 47.59 GB | 11.5 t/s |
| Apple M3 Max 96GB Apple | 96 GB | Runs well | Q8_0 | 47.59 GB | 11.5 t/s |
| Apple M3 Max 128GB Apple | 128 GB | Runs well | Q8_0 | 47.59 GB | 11.5 t/s |
| NVIDIA DGX Spark 128GB NVIDIA | 128 GB | Runs well | Q6_K | 37.07 GB | 10.8 t/s |
| GeForce RTX 4090 NVIDIA | 24 GB | Just fits | Q3_K_M | 22.6 GB | 62.8 t/s |
| GeForce RTX 3090 Ti NVIDIA | 24 GB | Just fits | Q3_K_M | 22.6 GB | 62.8 t/s |
| GeForce RTX 3090 NVIDIA | 24 GB | Just fits | Q3_K_M | 22.6 GB | 58.3 t/s |
Show 35 more devices
| GeForce RTX 5090 Laptop NVIDIA | 24 GB | Just fits | Q3_K_M | 22.6 GB | 55.8 t/s |
| Radeon RX 7900 XTX AMD | 24 GB | Just fits | Q3_K_M | 22.6 GB | 50.7 t/s |
| RTX A5000 NVIDIA | 24 GB | Just fits | Q3_K_M | 22.6 GB | 47.9 t/s |
| Apple M4 Pro 64GB Apple | 64 GB | Fits, but slow | Q6_K | 37.04 GB | 10 t/s |
| Jetson AGX Orin 64GB NVIDIA | 64 GB | Runs well | Q4_K_M | 27.66 GB | 10.6 t/s |
| Ryzen AI Max+ 395 64GB AMD | 64 GB | Runs well | Q4_K_M | 27.66 GB | 10.3 t/s |
| Ryzen AI Max+ 395 96GB AMD | 96 GB | Runs well | Q4_K_M | 27.66 GB | 10.3 t/s |
| Ryzen AI Max+ 395 128GB AMD | 128 GB | Runs well | Q4_K_M | 27.66 GB | 10.3 t/s |
| GeForce RTX 4060 Ti 16GB NVIDIA | 16 GB | Runs great | IQ2_XXS ! | 12.57 GB | 30.1 t/s |
| Radeon RX 7600 XT AMD | 16 GB | Runs great | IQ2_XXS ! | 12.57 GB | 25.5 t/s |
| Apple M3 Pro 18GB Apple | 18 GB | Just fits | IQ2_XXS ! | 12.57 GB | 14.5 t/s |
| Apple M4 Max 36GB Apple | 36 GB | Just fits | n/a | 32.31 GB | 22.6 t/s |
| Apple M4 Max 48GB Apple | 48 GB | Just fits | n/a | 37.04 GB | 19.9 t/s |
| Apple M1 Pro 16GB Apple | 16 GB | Just fits | n/a ! | 12.57 GB | 19.3 t/s |
| Apple M2 Pro 16GB Apple | 16 GB | Just fits | n/a ! | 12.57 GB | 19.3 t/s |
| Apple M4 Pro 24GB Apple | 24 GB | Just fits | n/a ! | 18.01 GB | 19.3 t/s |
| Apple M1 Max 32GB Apple | 32 GB | Just fits | n/a | 27.63 GB | 19.2 t/s |
| Apple M2 Max 32GB Apple | 32 GB | Just fits | n/a | 27.63 GB | 19.2 t/s |
| Apple M3 Max 36GB Apple | 36 GB | Just fits | n/a | 32.31 GB | 16.6 t/s |
| Apple M5 16GB Apple | 16 GB | Just fits | n/a ! | 12.57 GB | 14.8 t/s |
| Apple M3 Max 48GB Apple | 48 GB | Just fits | n/a | 37.04 GB | 14.6 t/s |
| Apple M4 16GB Apple | 16 GB | Just fits | n/a ! | 12.57 GB | 11.6 t/s |
| Intel Core Ultra 9 288V 16GB Intel | 16 GB | Just fits | n/a ! | 12.6 GB | 11 t/s |
| Apple M5 24GB Apple | 24 GB | Just fits | n/a ! | 18.01 GB | 10.8 t/s |
| Apple M1 Pro 32GB Apple | 32 GB | Just fits | n/a | 24.48 GB | 10.7 t/s |
| Apple M2 Pro 32GB Apple | 32 GB | Just fits | n/a | 24.48 GB | 10.7 t/s |
| Jetson AGX Orin 32GB NVIDIA | 32 GB | Just fits | n/a | 27.66 GB | 10.6 t/s |
| Ryzen AI Max+ 395 32GB AMD | 32 GB | Just fits | n/a | 27.66 GB | 10.3 t/s |
| Ryzen AI 9 HX 370 16GB AMD | 16 GB | Just fits | n/a ! | 12.6 GB | 10.3 t/s |
| Apple M4 Pro 48GB Apple | 48 GB | Just fits | n/a | 37.04 GB | 10 t/s |
| Apple M2 16GB Apple | 16 GB | Just fits | n/a ! | 12.57 GB | 9.7 t/s |
| Apple M3 16GB Apple | 16 GB | Just fits | n/a ! | 12.57 GB | 9.7 t/s |
| Apple M4 24GB Apple | 24 GB | Just fits | n/a ! | 18.01 GB | 8.5 t/s |
| Ryzen AI 9 HX 370 24GB AMD | 24 GB | Just fits | n/a ! | 18.04 GB | 7.6 t/s |
| Apple M5 32GB Apple | 32 GB | Just fits | n/a | 27.63 GB | 7.3 t/s |
How it compares
Other models in the same family or of a similar size, on a 12 GB card and a 24 GB card, each at the best quantization that fits.
| Model | Weights at 4 bits | GeForce RTX 3060 12GB | GeForce RTX 4090 |
|---|---|---|---|
| Mixtral 8x7B Instruct V0.1 | 26.3 GB | does not fit well | 62.8 t/s at Q3_K_M |
| Mistral Small 24B Instruct 2501 | 13.3 GB | does not fit well | 39.6 t/s at Q6_K |
| Codestral 22B V0.1 | 12.5 GB | does not fit well | 39.9 t/s at Q6_K |
| Mistral Nemo Instruct 2407 | 6.9 GB | 26.2 t/s at Q6_K | 58.4 t/s at Q8_0 |
| Gemma 4 31B Instruct | 17.6 GB | does not fit well | 35.5 t/s at Q4_K_M |
| Qwen3 32B | 18.4 GB | does not fit well | 37.1 t/s at Q4_K_M |
What to buy to run Mixtral 8x7B Instruct V0.1
The cheapest hardware that runs it at a quantization worth using and a speed you would not resent, at an 8k context window.
Cheapest that works
Radeon RX 7900 XT
$650
IQ4_XS · uses 24.45 GB of its 20 GB · about 15.5 tokens/s
Best value
Apple M1 Max 32GB
$1,400 whole machine
Q4_K_M · about 19.2 tokens/s · 400 GB/s
If it just has to run
GeForce GTX 1650
$100 used
Q4_K_M · about 4.1 tokens/s. Cheaper, and slower to answer.
| Device | Price | Memory | Runs it at | Speed | |
|---|---|---|---|---|---|
| Radeon RX 7900 XT AMD | $650 | 20 GB | IQ4_XS | 15.5 t/s | Check price |
| GeForce RTX 3090 NVIDIA | $700 used | 24 GB | Q4_K_M | 18.4 t/s | Check price |
| GeForce RTX 3090 Ti NVIDIA | $800 used | 24 GB | Q4_K_M | 18.8 t/s | Check price |
| Apple M1 Max 32GB Apple · whole machine | $1,400 | 32 GB | Q4_K_M | 19.2 t/s | Check price |
| Apple M1 Max 64GB Apple · whole machine | $1,900 | 64 GB | Q5_K_M | 16.6 t/s | Check price |
| GeForce RTX 5090 NVIDIA | $2,200 | 32 GB | Q4_K_M | 93 t/s | Check price |
| Apple M1 Ultra 128GB Apple · whole machine | $3,400 | 128 GB | Q8_0 | 23 t/s | Check price |
| Apple M3 Ultra 256GB Apple · whole machine | $5,599 | 256 GB | Q8_0 | 23.6 t/s | Check price |
Indicative prices reviewed 2026-09-01; used prices are marketplace typical. Options that cost more than a cheaper one with no more memory and no more speed are hidden. Price links are Amazon affiliate links. Change the standard, the context window or the budget in the buying tool.
Where it sits in the catalog
Source: mistralai/Mixtral-8x7B-Instruct-v0.1 on Hugging Face. Downloaded 298k times in the last month. Published 2023-12-10. Architecture figures are read from the repository's own configuration file, so they move when the model does.
Questions people ask about Mixtral 8x7B Instruct V0.1
How much VRAM does Mixtral 8x7B Instruct V0.1 need?
At Q4_K_M with an 8k context window, Mixtral 8x7B Instruct V0.1 needs about 27.59 GB: 26.19 GB of weights, 1 GB of KV cache and 0.41 GB of overhead. At the highest-quality file it needs about 47.49 GB.
What is the cheapest GPU that runs Mixtral 8x7B Instruct V0.1?
The smallest consumer device in our catalog that fits it is the GeForce RTX 4090 with 24 GB, running it at Q3_K_M at roughly 62.8 tokens per second.
How much does a longer context window add for Mixtral 8x7B Instruct V0.1?
On the memory side, a 32k window adds 4.0 GB of KV cache on top of the weights, and the cache grows in a straight line with every token. The longest window the model supports natively is 32k tokens.