DeepSeek on AMD RDNA3

Can I run DeepSeek-R1-Distill-Qwen 32B on an Radeon RX 7900 XTX?

Yes. DeepSeek-R1-Distill-Qwen 32B at Q4_K_M uses 21.5 GB of the 21.8 GB available on a Radeon RX 7900 XTX, and runs at about 27.6 tokens per second. You can push the context to 9K.

Fits, but tight

21.5 GB of 21.8 GB · 99%
022 GB
Weights 18.6 GB
KV cache 2.0 GB
Runtime overhead 0.8 GB

This fits with almost nothing to spare. A background application claiming VRAM will push it over. Drop to the next quantisation down, or quantise the KV cache to Q8_0 — that halves the cache for no meaningful quality loss.

Generation27.6tok/s
Prompt processing788tok/s
Max context9Ktokens
KV per 1K tokens0GB

Every quantisation of DeepSeek-R1-Distill-Qwen 32B on a Radeon RX 7900 XTX

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

QuantisationbpwWeights TotalVerdictMax ctxtok/s
FP16 / BF16 16.00 61.0 GB 63.9 GB Won't fit 0.81
INT8 / W8A8 8.50 32.1 GB 34.9 GB Won't fit 2.38
Q8_0 (GGUF) 8.50 32.1 GB 34.9 GB Won't fit 2.38
FP8 (E4M3) 8.00 30.5 GB 33.3 GB Won't fit 2.67
Q6_K 6.56 25.0 GB 27.9 GB Won't fit 4.65
Q5_K_M 5.67 21.6 GB 24.5 GB Won't fit 8.73
Q5_K_S 5.52 21.1 GB 23.9 GB Won't fit 9.73
Q4_K_M 4.85 18.6 GB 21.5 GB Fits, but tight 9K 27.6
AWQ 4-bit 4.25 18.3 GB 21.2 GB Fits, but tight 10K 27.9
GPTQ 4-bit 4.25 18.3 GB 21.2 GB Fits, but tight 10K 27.9
MXFP4 4.25 18.3 GB 21.2 GB Fits, but tight 10K 27.9
Q4_K_S 4.58 17.6 GB 20.5 GB Fits, but tight 13K 29.0
Q4_0 4.55 17.5 GB 20.4 GB Fits, but tight 14K 29.2
IQ4_XS 4.25 16.4 GB 19.3 GB Runs comfortably 18K 31.0
Q3_K_M 3.91 15.2 GB 18.0 GB Runs comfortably 23K 33.3
IQ3_M 3.70 14.4 GB 17.3 GB Runs comfortably 26K 34.9
IQ3_XXS 3.06 12.1 GB 15.0 GB Runs comfortably 35K 41.0
Q2_K 2.63 10.6 GB 13.4 GB Runs comfortably 41K 46.5
IQ2_XXS 2.06 8.5 GB 11.3 GB Runs comfortably 50K 56.4
IQ1_M 1.75 7.4 GB 10.2 GB Runs comfortably 54K 63.8

Also worth checking