run it locally

Run Mixtral 8x7B on A100 80GB

46.7B parameters on a 80 GB card. It fits at Q8_0 — here's the VRAM math at every quantization, the speed to expect, and the quality you trade.

verdict · Mixtral 8x7B on A100 80GB
Fits at Q8_0
needs 59.4 GB at Q8_0 · 76.0 GB usable · ~35 tokens/sec
VRAM needed and fit for Mixtral 8x7B on A100 80GB by quantization.
quantization vram needed fits 80gb? tokens/sec quality
FP16 (full) 112.1 GB ✗ no Reference quality — no quantization loss.
Q8_0recommended 59.4 GB ✓ yes ~35 Near-lossless; rarely worth the extra space over Q6.
Q6_K 46.0 GB ✓ yes ~45 Virtually indistinguishable from full precision.
Q5_K_M 38.7 GB ✓ yes ~54 Minor loss; an excellent quality-vs-size balance.
Q4_K_M 31.4 GB ✓ yes ~66 Small but measurable loss; the popular default.
Q3_K_M 24.1 GB ✓ yes ~86 Noticeable degradation; only when you're tight on VRAM.

Weights = params × bytes/weight, +20% for KV cache & runtime; usable VRAM is 95% of nameplate. Tokens/sec is a bandwidth ceiling (2039 GB/s) — real throughput is lower with long context. Try other combinations →

Running Mixtral 8x7B on a A100 80GB

At Q8_0, Mixtral 8x7B's weights take about 59 GB, inside the A100 80GB's 76 GB of usable memory. Near-lossless; rarely worth the extra space over Q6. Generation speed is bound by memory bandwidth — the GPU reads the whole model once per token — so expect on the order of 35 tokens/sec before context overhead.

Quantization is the lever

Each step down in precision shrinks the model: FP16 needs about 112 GB, Q4_K_M about 31 GB — a 72% reduction for a small, usually acceptable quality cost. Q5_K_M and Q6_K are near-lossless if you have the headroom; drop to Q3 only when you're genuinely out of VRAM. The quantization guide covers the tradeoffs in detail.

Frequently asked questions

Can a A100 80GB run Mixtral 8x7B?

Yes — Mixtral 8x7B fits on a A100 80GB at Q8_0 (59 GB of the 76 GB usable), with roughly 35 tokens/sec. Higher-precision quants need more VRAM; the table above shows each option.

How much VRAM does Mixtral 8x7B need?

Mixtral 8x7B is a 46.7B-parameter model. At FP16 that's about 112 GB; at Q4_K_M (the popular default) about 31 GB, including ~20% for the KV cache and runtime. Quantization is the main lever — see the per-quant table above.

Other combinations

Mixtral 8x7B on other GPUs: RTX 3060 12GB, RTX 4070 Ti, RTX 4080, RTX 3090, RTX 4090, NVIDIA L4

Other models on the A100 80GB: Qwen2.5 32B, Command R 35B, Llama 3.3 70B, Qwen2.5 72B, Mixtral 8x22B

Related