Qwen on NVIDIA Ampere

Can I run Qwen3 14B on an A100 80GB?

Yes. Qwen3 14B at Q4_K_M uses 10.5 GB of the 74.4 GB available on a A100 80GB, and runs at about 136 tokens per second. You can push the context to 128K.

Runs comfortably

10.5 GB of 74.4 GB · 14%
074 GB
Weights 8.5 GB
KV cache 1.3 GB
Runtime overhead 0.8 GB

Qwen3 14B at Q4_K_M leaves 63.9 GB spare on a A100 80GB. There is room to raise the context length or move up a quantisation level.

Generation136tok/s
Prompt processing4436tok/s
Max context128Ktokens
KV per 1K tokens0GB

Every quantisation of Qwen3 14B on a A100 80GB

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

QuantisationbpwWeights TotalVerdictMax ctxtok/s
FP16 / BF16 16.00 27.5 GB 29.6 GB Runs comfortably 128K 44.9
INT8 / W8A8 8.50 14.3 GB 16.4 GB Runs comfortably 128K 84.0
Q8_0 (GGUF) 8.50 14.3 GB 16.4 GB Runs comfortably 128K 84.0
FP8 (E4M3) 8.00 13.8 GB 15.8 GB Runs comfortably 128K 86.9
Q6_K 6.56 11.3 GB 13.4 GB Runs comfortably 128K 105
Q5_K_M 5.67 9.7 GB 11.8 GB Runs comfortably 128K 119
Q5_K_S 5.52 9.5 GB 11.6 GB Runs comfortably 128K 122
AWQ 4-bit 4.25 9.4 GB 11.5 GB Runs comfortably 128K 123
GPTQ 4-bit 4.25 9.4 GB 11.5 GB Runs comfortably 128K 123
MXFP4 4.25 9.4 GB 11.5 GB Runs comfortably 128K 123
Q4_K_M 4.85 8.5 GB 10.5 GB Runs comfortably 128K 136
Q4_K_S 4.58 8.0 GB 10.1 GB Runs comfortably 128K 142
Q4_0 4.55 8.0 GB 10.1 GB Runs comfortably 128K 143
IQ4_XS 4.25 7.5 GB 9.6 GB Runs comfortably 128K 151
Q3_K_M 3.91 7.0 GB 9.1 GB Runs comfortably 128K 161
IQ3_M 3.70 6.7 GB 8.8 GB Runs comfortably 128K 168
IQ3_XXS 3.06 5.7 GB 7.8 GB Runs comfortably 128K 192
Q2_K 2.63 5.0 GB 7.1 GB Runs comfortably 128K 214
IQ2_XXS 2.06 4.2 GB 6.3 GB Runs comfortably 128K 250
IQ1_M 1.75 3.7 GB 5.8 GB Runs comfortably 128K 276

Also worth checking

Qwen3 14B on every GPU
Full hardware tableEverything that runs on a A100 80GB
Full model tableWhy the KV cache is the part that bites
Guide