How much VRAM does Mixtral 8x7B need?
Mixtral 8x7B needs about 28.4 GB of VRAM at Q4_K_M, or 49.3 GB at Q8_0 and 92.1 GB at FP16. No consumer card holds it at Q4_K_M. The smallest fit in our set is the RTX 5090 (32 GB). Weights plus runtime overhead; KV cache for your context length is on top.
Weights-plus-runtime footprint across the six most common quantizations. KV cache is context-dependent and comes on top. At 32K it is gigabytes on its own, before the weights; size it precisely in the GPU Memory Calculator.
This is a mixture-of-experts model: all 46.7B parameters must sit in memory, but only about 12.9B are active per token — so it loads like a 46.7B model and runs closer to a 12.9B one.
This is the base checkpoint: pretrained on raw text and not tuned to follow instructions, so it continues text rather than answering prompts. That makes it the right starting point if you plan to fine-tune your own model. For chat or assistant work, use Mixtral 8x7B Instruct — same parameter count, so the memory figures below are identical.
Can your GPU run Mixtral 8x7B?
Pick your card — the answer below is computed by the same engine that produces every figure on this page.
A100 40GB runs Mixtral 8x7B at Q4_K_M with 9.6 GB to spare; room left for context and everyday runtime pressure; as a mixture-of-experts model all 46.7B parameters must be resident even though only 12.9B are active per token.
Weights plus runtime overhead. KV cache is not included — it depends on context length and this model’s published architecture.
| Weights (Q4_K_M) | 26.4 GB |
| Runtime overhead | 2.1 GB |
| Estimated total | 28.4 GB |
| Usable VRAM (95% of 40 GB) | 38.0 GB |
Across every GPU we track
- RTX 3060 12GBNot enough-17.0 GB
- RTX 4060 Ti 16GBNot enough-13.2 GB
- RTX 4070 Ti SUPERNot enough-13.2 GB
- RTX 3090Not enough-5.6 GB
- RTX 4090Not enough-5.6 GB
- RTX 5090Tight+2.0 GB
- A100 40GBRecommended+9.6 GB
- RTX A6000Recommended+17.2 GB
- L40SRecommended+17.2 GB
- A100 80GBExcellent+47.6 GB
- H100 80GBExcellent+47.6 GB
- H200 141GBExcellent+105.5 GB
Estimates, not guarantees: real usage moves with runtime, driver, batch size and context. Size a specific context window in the GPU Memory Calculator.
How was this number calculated? Every figure here comes from Bitpute’s documented calculation methodology — parameters, precision, quantization, runtime overhead and usable VRAM, each formula written out in full.
Memory by quantization
| Quant | Weights | + runtime overhead |
|---|---|---|
| FP16 | 87.0 GB | 92.1 GB |
| Q8_0 | 46.2 GB | 49.3 GB |
| Q6_K | 35.7 GB | 38.2 GB |
| Q5_K_M | 30.9 GB | 33.2 GB |
| Q4_K_M | 26.4 GB | 28.4 GB |
| Q4_0 | 24.5 GB | 26.4 GB |
Overhead = 0.75 GB + 5% of weights (CUDA context, buffers). Add KV cache on top. Weights dominate at this size, but a 32K window still costs gigabytes.
Which quantization should you actually run?
For Mixtral 8x7B, the single-GPU entry point is Q4_K_M (28.4 GB) on RTX A6000, with plenty of headroom for long context. Higher quality (Q6_K 38.2 GB, FP16 92.1 GB) needs a 48 GB+ card or multiple GPUs — for occasional use, renting is usually cheaper than buying. As a mixture-of-experts model it must hold all 46.7B in VRAM but activates only about 12.9B per token, so it runs much faster than its footprint suggests — memory is the limit here, not speed.
Single-GPU fit at Q4_K_M (28.4 GB + KV)
RTX 3060 12GBRTX 4060 Ti 16GBRTX 4070 Ti SUPERRTX 3090RTX 4090RTX 5090A100 40GBRTX A6000L40SA100 80GBH100 80GBH200 141GB
Where the memory goes
Weights and overhead are exact for Q4_K_M. The KV bar is a generic 32-layer transformer at 8K; at this scale the weights dominate, but long context still adds gigabytes — size it in the calculator is for.
Deciding how much headroom to buy at this tier? RTX 5090 vs A100 40GB compares them on bandwidth, power and which models each one holds.
Common questions
How much VRAM does Mixtral 8x7B need?
Mixtral 8x7B needs about 28.4 GB of VRAM at Q4_K_M, 49.3 GB at Q8_0, or 92.1 GB at FP16. That is weights plus about 0.75 GB of runtime overhead and 5% of weight size; KV cache is additional and depends on context length.
What GPU can run Mixtral 8x7B?
At Q4_K_M the smallest card in our set that fits is the RTX 5090 (32 GB usable at 95%). At FP16 you need the H200 141GB (141 GB) or larger.
Can Mixtral 8x7B run on 24 GB?
Not at Q4_K_M: it needs about 28.4 GB, more than the ~22.8 GB usable on a 24 GB card. Use a larger card, multiple GPUs, or offload.
Same memory footprint as 1 other checkpoint
Mixtral 8x7B has an identical parameter count to this, so every figure on this page applies to it unchanged. What varies is different checkpoint roles (instruction-tuned) — not the size. Choose on capability and licence; the memory budget is identical.