Mistral on NVIDIA Ada

Can I run Mistral NeMo 12B on an RTX 4070?

Yes. Mistral NeMo 12B at Q4_K_M uses 9.1 GB of the 10.5 GB available on a RTX 4070, and runs at about 39.2 tokens per second. You can push the context to 17K.

Runs comfortably

9.1 GB of 10.5 GB · 87%
010 GB
Weights 7.0 GB
KV cache 1.3 GB
Runtime overhead 0.8 GB

Mistral NeMo 12B at Q4_K_M leaves 1.4 GB spare on a RTX 4070. There is room to raise the context length or move up a quantisation level.

Generation39.2tok/s
Prompt processing994tok/s
Max context17Ktokens
KV per 1K tokens0GB

Every quantisation of Mistral NeMo 12B on a RTX 4070

Highlighted row is the highest quality that still fits at 8K context.

QuantisationbpwWeights TotalVerdictMax ctxtok/s
FP16 / BF16 16.00 22.8 GB 24.9 GB Won't fit — 2.35
INT8 / W8A8 8.50 11.8 GB 13.9 GB Won't fit — 7.90
Q8_0 (GGUF) 8.50 11.8 GB 13.9 GB Won't fit — 7.90
FP8 (E4M3) 8.00 11.4 GB 13.5 GB Won't fit — 8.65
Q6_K 6.56 9.4 GB 11.4 GB Won't fit 2K 16.2
Q5_K_M 5.67 8.1 GB 10.2 GB Fits, but tight 10K 34.4
AWQ 4-bit 4.25 7.9 GB 10.0 GB Fits, but tight 11K 35.2
GPTQ 4-bit 4.25 7.9 GB 10.0 GB Fits, but tight 11K 35.2
MXFP4 4.25 7.9 GB 10.0 GB Fits, but tight 11K 35.2
Q5_K_S 5.52 7.9 GB 10.0 GB Fits, but tight 11K 35.3
Q4_K_M 4.85 7.0 GB 9.1 GB Runs comfortably 17K 39.2
Q4_K_S 4.58 6.7 GB 8.8 GB Runs comfortably 19K 41.0
Q4_0 4.55 6.6 GB 8.7 GB Runs comfortably 19K 41.2
IQ4_XS 4.25 6.3 GB 8.3 GB Runs comfortably 22K 43.5
Q3_K_M 3.91 5.8 GB 7.9 GB Runs comfortably 24K 46.3
IQ3_M 3.70 5.6 GB 7.6 GB Runs comfortably 26K 48.3
IQ3_XXS 3.06 4.7 GB 6.8 GB Runs comfortably 31K 55.5
Q2_K 2.63 4.2 GB 6.3 GB Runs comfortably 35K 61.7
IQ2_XXS 2.06 3.5 GB 5.6 GB Runs comfortably 39K 72.4
IQ1_M 1.75 3.1 GB 5.2 GB Runs comfortably 42K 80.0

Also worth checking

Mistral NeMo 12B on every GPU
Full hardware tableEverything that runs on a RTX 4070
Full model tableWhy the KV cache is the part that bites
Guide