Mistral on NVIDIA Ada

Can I run Mistral Small 24B on an L40S?

Yes. Mistral Small 24B at Q4_K_M uses 15.5 GB of the 44.3 GB available on a L40S, and runs at about 36.6 tokens per second. You can push the context to 32K.

Runs comfortably

15.5 GB of 44.3 GB · 35%
044 GB
Weights 13.4 GB
KV cache 1.3 GB
Runtime overhead 0.8 GB

Mistral Small 24B at Q4_K_M leaves 28.8 GB spare on a L40S. There is room to raise the context length or move up a quantisation level.

Generation36.6tok/s
Prompt processing1613tok/s
Max context32Ktokens
KV per 1K tokens0GB

Every quantisation of Mistral Small 24B on a L40S

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

QuantisationbpwWeights TotalVerdictMax ctxtok/s
FP16 / BF16 16.00 43.9 GB 46.0 GB Won't fit — 7.14
INT8 / W8A8 8.50 23.0 GB 25.1 GB Runs comfortably 32K 21.8
Q8_0 (GGUF) 8.50 23.0 GB 25.1 GB Runs comfortably 32K 21.8
FP8 (E4M3) 8.00 22.0 GB 24.0 GB Runs comfortably 32K 22.8
Q6_K 6.56 18.0 GB 20.1 GB Runs comfortably 32K 27.6
Q5_K_M 5.67 15.6 GB 17.6 GB Runs comfortably 32K 31.8
Q5_K_S 5.52 15.1 GB 17.2 GB Runs comfortably 32K 32.6
AWQ 4-bit 4.25 13.5 GB 15.6 GB Runs comfortably 32K 36.4
GPTQ 4-bit 4.25 13.5 GB 15.6 GB Runs comfortably 32K 36.4
MXFP4 4.25 13.5 GB 15.6 GB Runs comfortably 32K 36.4
Q4_K_M 4.85 13.4 GB 15.5 GB Runs comfortably 32K 36.6
Q4_K_S 4.58 12.7 GB 14.8 GB Runs comfortably 32K 38.5
Q4_0 4.55 12.6 GB 14.7 GB Runs comfortably 32K 38.7
IQ4_XS 4.25 11.9 GB 13.9 GB Runs comfortably 32K 41.1
Q3_K_M 3.91 11.0 GB 13.1 GB Runs comfortably 32K 44.2
IQ3_M 3.70 10.4 GB 12.5 GB Runs comfortably 32K 46.3
IQ3_XXS 3.06 8.8 GB 10.9 GB Runs comfortably 32K 54.4
Q2_K 2.63 7.7 GB 9.8 GB Runs comfortably 32K 61.6
IQ2_XXS 2.06 6.2 GB 8.3 GB Runs comfortably 32K 74.7
IQ1_M 1.75 5.4 GB 7.5 GB Runs comfortably 32K 84.4

Also worth checking

Mistral Small 24B on every GPU
Full hardware tableEverything that runs on a L40S
Full model tableWhy the KV cache is the part that bites
Guide