Mistral on NVIDIA Ampere

Can I run Mistral 7B v0.3 on an RTX 3080 10GB?

Yes. Mistral 7B v0.3 at Q4_K_M uses 5.9 GB of the 8.6 GB available on a RTX 3080 10GB, and runs at about 101 tokens per second. You can push the context to 29K.

Runs comfortably

5.9 GB of 8.6 GB · 69%
09 GB
Weights 4.1 GB
KV cache 1.0 GB
Runtime overhead 0.8 GB

Mistral 7B v0.3 at Q4_K_M leaves 2.7 GB spare on a RTX 3080 10GB. There is room to raise the context length or move up a quantisation level.

Generation101tok/s
Prompt processing1738tok/s
Max context29Ktokens
KV per 1K tokens0GB

Every quantisation of Mistral 7B v0.3 on a RTX 3080 10GB

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

QuantisationbpwWeights TotalVerdictMax ctxtok/s
FP16 / BF16 16.00 13.5 GB 15.3 GB Won't fit 5.26
INT8 / W8A8 8.50 7.1 GB 8.9 GB Won't fit 5K 36.9
Q8_0 (GGUF) 8.50 7.1 GB 8.9 GB Won't fit 5K 36.9
FP8 (E4M3) 8.00 6.8 GB 8.6 GB Fits, but tight 8K 64.8
Q6_K 6.56 5.5 GB 7.4 GB Runs comfortably 18K 77.6
Q5_K_M 5.67 4.8 GB 6.6 GB Runs comfortably 24K 88.5
Q5_K_S 5.52 4.7 GB 6.5 GB Runs comfortably 25K 90.6
Q4_K_M 4.85 4.1 GB 5.9 GB Runs comfortably 29K 101
AWQ 4-bit 4.25 4.0 GB 5.8 GB Runs comfortably 31K 105
GPTQ 4-bit 4.25 4.0 GB 5.8 GB Runs comfortably 31K 105
MXFP4 4.25 4.0 GB 5.8 GB Runs comfortably 31K 105
Q4_K_S 4.58 3.9 GB 5.7 GB Runs comfortably 31K 106
Q4_0 4.55 3.9 GB 5.7 GB Runs comfortably 31K 107
IQ4_XS 4.25 3.6 GB 5.4 GB Runs comfortably 32K 113
Q3_K_M 3.91 3.3 GB 5.2 GB Runs comfortably 32K 121
IQ3_M 3.70 3.2 GB 5.0 GB Runs comfortably 32K 126
IQ3_XXS 3.06 2.7 GB 4.5 GB Runs comfortably 32K 146
Q2_K 2.63 2.3 GB 4.1 GB Runs comfortably 32K 164
IQ2_XXS 2.06 1.8 GB 3.7 GB Runs comfortably 32K 195
IQ1_M 1.75 1.6 GB 3.4 GB Runs comfortably 32K 218

Also worth checking