Qwen on AMD Strix Halo

Can I run Qwen3 14B on an Ryzen AI Max+ 395 128GB?

Yes. Qwen3 14B at Q4_K_M uses 10.2 GB of the 96.0 GB available on a Ryzen AI Max+ 395 128GB, and runs at about 14.8 tokens per second. You can push the context to 128K.

Runs comfortably

10.2 GB of 96.0 GB · 11%
096 GB
Weights 8.5 GB
KV cache 1.3 GB
Runtime overhead 0.5 GB

Qwen3 14B at Q4_K_M leaves 85.8 GB spare on a Ryzen AI Max+ 395 128GB. There is room to raise the context length or move up a quantisation level.

Generation14.8tok/s
Prompt processing711tok/s
Max context128Ktokens
KV per 1K tokens0GB

Every quantisation of Qwen3 14B on a Ryzen AI Max+ 395 128GB

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

QuantisationbpwWeights TotalVerdictMax ctxtok/s
FP16 / BF16 16.00 27.5 GB 29.3 GB Runs comfortably 128K 4.77
INT8 / W8A8 8.50 14.3 GB 16.1 GB Runs comfortably 128K 9.01
Q8_0 (GGUF) 8.50 14.3 GB 16.1 GB Runs comfortably 128K 9.01
FP8 (E4M3) 8.00 13.8 GB 15.5 GB Runs comfortably 128K 9.33
Q6_K 6.56 11.3 GB 13.1 GB Runs comfortably 128K 11.3
Q5_K_M 5.67 9.7 GB 11.5 GB Runs comfortably 128K 12.9
Q5_K_S 5.52 9.5 GB 11.3 GB Runs comfortably 128K 13.2
AWQ 4-bit 4.25 9.4 GB 11.2 GB Runs comfortably 128K 13.3
GPTQ 4-bit 4.25 9.4 GB 11.2 GB Runs comfortably 128K 13.3
MXFP4 4.25 9.4 GB 11.2 GB Runs comfortably 128K 13.3
Q4_K_M 4.85 8.5 GB 10.2 GB Runs comfortably 128K 14.8
Q4_K_S 4.58 8.0 GB 9.8 GB Runs comfortably 128K 15.5
Q4_0 4.55 8.0 GB 9.8 GB Runs comfortably 128K 15.5
IQ4_XS 4.25 7.5 GB 9.3 GB Runs comfortably 128K 16.4
Q3_K_M 3.91 7.0 GB 8.8 GB Runs comfortably 128K 17.5
IQ3_M 3.70 6.7 GB 8.5 GB Runs comfortably 128K 18.3
IQ3_XXS 3.06 5.7 GB 7.5 GB Runs comfortably 128K 21.1
Q2_K 2.63 5.0 GB 6.8 GB Runs comfortably 128K 23.6
IQ2_XXS 2.06 4.2 GB 6.0 GB Runs comfortably 128K 27.9
IQ1_M 1.75 3.7 GB 5.5 GB Runs comfortably 128K 30.9

Also worth checking

Qwen3 14B on every GPU
Full hardware tableEverything that runs on a Ryzen AI Max+ 395 128GB
Full model tableWhy the KV cache is the part that bites
Guide