Qwen on NVIDIA Ampere

Can I run Qwen3 4B on an RTX A6000?

Yes. Qwen3 4B at Q4_K_M uses 4.2 GB of the 44.3 GB available on a RTX A6000, and runs at about 162 tokens per second. You can push the context to 128K.

Runs comfortably

4.2 GB of 44.3 GB · 10%
044 GB
Weights 2.3 GB
KV cache 1.1 GB
Runtime overhead 0.8 GB

Qwen3 4B at Q4_K_M leaves 40.1 GB spare on a RTX A6000. There is room to raise the context length or move up a quantisation level.

Generation162tok/s
Prompt processing3918tok/s
Max context128Ktokens
KV per 1K tokens0GB

Every quantisation of Qwen3 4B on a RTX A6000

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

QuantisationbpwWeights TotalVerdictMax ctxtok/s
FP16 / BF16 16.00 7.5 GB 9.4 GB Runs comfortably 128K 59.0
INT8 / W8A8 8.50 3.9 GB 5.8 GB Runs comfortably 128K 106
Q8_0 (GGUF) 8.50 3.9 GB 5.8 GB Runs comfortably 128K 106
FP8 (E4M3) 8.00 3.7 GB 5.7 GB Runs comfortably 128K 109
Q6_K 6.56 3.1 GB 5.0 GB Runs comfortably 128K 129
Q5_K_M 5.67 2.7 GB 4.6 GB Runs comfortably 128K 145
Q5_K_S 5.52 2.6 GB 4.5 GB Runs comfortably 128K 148
AWQ 4-bit 4.25 2.5 GB 4.4 GB Runs comfortably 128K 151
GPTQ 4-bit 4.25 2.5 GB 4.4 GB Runs comfortably 128K 151
MXFP4 4.25 2.5 GB 4.4 GB Runs comfortably 128K 151
Q4_K_M 4.85 2.3 GB 4.2 GB Runs comfortably 128K 162
Q4_K_S 4.58 2.2 GB 4.1 GB Runs comfortably 128K 169
Q4_0 4.55 2.2 GB 4.1 GB Runs comfortably 128K 169
IQ4_XS 4.25 2.0 GB 4.0 GB Runs comfortably 128K 177
Q3_K_M 3.91 1.9 GB 3.8 GB Runs comfortably 128K 187
IQ3_M 3.70 1.8 GB 3.7 GB Runs comfortably 128K 194
IQ3_XXS 3.06 1.5 GB 3.5 GB Runs comfortably 128K 218
Q2_K 2.63 1.4 GB 3.3 GB Runs comfortably 128K 237
IQ2_XXS 2.06 1.1 GB 3.0 GB Runs comfortably 128K 269
IQ1_M 1.75 1.0 GB 2.9 GB Runs comfortably 128K 290

Also worth checking

Qwen3 4B on every GPU
Full hardware tableEverything that runs on a RTX A6000
Full model tableWhy the KV cache is the part that bites
Guide