run it locally
Run Mixtral 8x22B on Apple M2 Ultra 192GB
141B parameters on a 192 GB (unified memory) card. It fits at Q5_K_M — here's the VRAM math at every quantization, the speed to expect, and the quality you trade.
| quantization | vram needed | fits 192gb? | tokens/sec | quality |
|---|---|---|---|---|
| FP16 (full) | 338.4 GB | ✗ no | — | Reference quality — no quantization loss. |
| Q8_0 | 179.4 GB | ✗ no | — | Near-lossless; rarely worth the extra space over Q6. |
| Q6_K | 138.7 GB | ✗ no | — | Virtually indistinguishable from full precision. |
| Q5_K_Mrecommended | 116.7 GB | ✓ yes | ~7 | Minor loss; an excellent quality-vs-size balance. |
| Q4_K_M | 94.8 GB | ✓ yes | ~9 | Small but measurable loss; the popular default. |
| Q3_K_M | 72.8 GB | ✓ yes | ~11 | Noticeable degradation; only when you're tight on VRAM. |
Weights = params × bytes/weight, +20% for KV cache & runtime; usable VRAM is 72% (Apple unified memory) of nameplate. Tokens/sec is a bandwidth ceiling (800 GB/s) — real throughput is lower with long context. Try other combinations →
Running Mixtral 8x22B on a Apple M2 Ultra 192GB
At Q5_K_M, Mixtral 8x22B's weights take about 117 GB, inside the Apple M2 Ultra 192GB's 138 GB of usable memory. Minor loss; an excellent quality-vs-size balance. Generation speed is bound by memory bandwidth — the GPU reads the whole model once per token — so expect on the order of 7 tokens/sec before context overhead.
Quantization is the lever
Each step down in precision shrinks the model: FP16 needs about 338 GB, Q4_K_M about 95 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 Apple M2 Ultra 192GB run Mixtral 8x22B?
Yes — Mixtral 8x22B fits on a Apple M2 Ultra 192GB at Q5_K_M (117 GB of the 138 GB usable), with roughly 7 tokens/sec. Higher-precision quants need more VRAM; the table above shows each option.
How much VRAM does Mixtral 8x22B need?
Mixtral 8x22B is a 141B-parameter model. At FP16 that's about 338 GB; at Q4_K_M (the popular default) about 95 GB, including ~20% for the KV cache and runtime. Quantization is the main lever — see the per-quant table above.
Other combinations
Mixtral 8x22B on other GPUs: RTX A6000, NVIDIA L40S, A100 40GB, A100 80GB, H100 80GB, H200 141GB
Other models on the Apple M2 Ultra 192GB: Llama 3.3 70B, Qwen2.5 72B, Llama 3.1 405B
Related
- VRAM calculator — any model, quant and GPU.
- Running models locally — the hardware reality.
- All run-locally combinations →