Mistral on NVIDIA Ada
Can I run Mixtral 8x7B on an RTX 4070?
Not at Q4_K_M — it needs 28.2 GB against 10.5 GB available. You would need 3 of these cards.
Won't fit
28.2 GB of 10.5 GB · 269%029 GB
Weights
26.4 GB
KV cache
1.0 GB
Runtime overhead
0.8 GB
Over the limit
17.7 GB
Short by 17.7 GB. You can run it with 10 of 32 layers on the RTX 4070 and the rest in system RAM, at roughly 6.67 tok/s — usable for batch work, painful for chat. A smaller quantisation or a shorter context is usually the better trade.
Generation6.67tok/s
Prompt processing944tok/s
Max context0tokens
KV per 1K tokens0GB
Every quantisation of Mixtral 8x7B on a RTX 4070
Highlighted row is the highest quality that still fits at 8K context.
| Quantisation | bpw | Weights | Total | Verdict | Max ctx | tok/s |
|---|---|---|---|---|---|---|
| FP16 / BF16 | 16.00 | 87.0 GB | 88.8 GB | Won't fit | — | 1.69 |
| INT8 / W8A8 | 8.50 | 46.2 GB | 48.0 GB | Won't fit | — | 3.43 |
| Q8_0 (GGUF) | 8.50 | 46.2 GB | 48.0 GB | Won't fit | — | 3.43 |
| FP8 (E4M3) | 8.00 | 43.5 GB | 45.3 GB | Won't fit | — | 3.63 |
| Q6_K | 6.56 | 35.7 GB | 37.5 GB | Won't fit | — | 4.53 |
| Q5_K_M | 5.67 | 30.8 GB | 32.6 GB | Won't fit | — | 5.38 |
| Q5_K_S | 5.52 | 30.0 GB | 31.8 GB | Won't fit | — | 5.71 |
| Q4_K_M | 4.85 | 26.4 GB | 28.2 GB | Won't fit | — | 6.67 |
| Q4_K_S | 4.58 | 24.9 GB | 26.7 GB | Won't fit | — | 7.30 |
| Q4_0 | 4.55 | 24.8 GB | 26.6 GB | Won't fit | — | 7.34 |
| AWQ 4-bit | 4.25 | 23.5 GB | 25.3 GB | Won't fit | — | 7.72 |
| GPTQ 4-bit | 4.25 | 23.5 GB | 25.3 GB | Won't fit | — | 7.72 |
| MXFP4 | 4.25 | 23.5 GB | 25.3 GB | Won't fit | — | 7.72 |
| IQ4_XS | 4.25 | 23.1 GB | 25.0 GB | Won't fit | — | 7.81 |
| Q3_K_M | 3.91 | 21.3 GB | 23.1 GB | Won't fit | — | 9.12 |
| IQ3_M | 3.70 | 20.2 GB | 22.0 GB | Won't fit | — | 9.58 |
| IQ3_XXS | 3.06 | 16.7 GB | 18.5 GB | Won't fit | — | 12.9 |
| Q2_K | 2.63 | 14.4 GB | 16.2 GB | Won't fit | — | 17.1 |
| IQ2_XXS | 2.06 | 11.3 GB | 13.1 GB | Won't fit | — | 28.7 |
| IQ1_M | 1.75 | 9.6 GB | 11.4 GB | Won't fit | — | 46.2 |