Qwen on NVIDIA Blackwell

Can I run Qwen2.5 72B on an RTX 5090?

Not at Q4_K_M — it needs 44.6 GB against 29.3 GB available. You would need 2 of these cards.

Won't fit

44.6 GB of 29.3 GB · 152%
046 GB
Weights 41.2 GB
KV cache 2.5 GB
Runtime overhead 0.9 GB
Over the limit 15.4 GB

Short by 15.4 GB. You can run it with 50 of 80 layers on the RTX 5090 and the rest in system RAM, at roughly 2.50 tok/s — usable for batch work, painful for chat. A smaller quantisation or a shorter context is usually the better trade.

Generation2.50tok/s
Prompt processing604tok/s
Max context0tokens
KV per 1K tokens0GB

Every quantisation of Qwen2.5 72B on a RTX 5090

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

QuantisationbpwWeights TotalVerdictMax ctxtok/s
FP16 / BF16 16.00 135.4 GB 138.8 GB Won't fit — 0.38
INT8 / W8A8 8.50 71.4 GB 74.8 GB Won't fit — 0.89
Q8_0 (GGUF) 8.50 71.4 GB 74.8 GB Won't fit — 0.89
FP8 (E4M3) 8.00 67.7 GB 71.1 GB Won't fit — 0.96
Q6_K 6.56 55.5 GB 58.9 GB Won't fit — 1.34
Q5_K_M 5.67 48.0 GB 51.4 GB Won't fit — 1.78
Q5_K_S 5.52 46.7 GB 50.1 GB Won't fit — 1.87
Q4_K_M 4.85 41.2 GB 44.6 GB Won't fit — 2.50
AWQ 4-bit 4.25 39.4 GB 42.8 GB Won't fit — 2.78
GPTQ 4-bit 4.25 39.4 GB 42.8 GB Won't fit — 2.78
MXFP4 4.25 39.4 GB 42.8 GB Won't fit — 2.78
Q4_K_S 4.58 39.0 GB 42.4 GB Won't fit — 2.90
Q4_0 4.55 38.8 GB 42.2 GB Won't fit — 2.91
IQ4_XS 4.25 36.3 GB 39.7 GB Won't fit — 3.45
Q3_K_M 3.91 33.6 GB 36.9 GB Won't fit — 4.57
IQ3_M 3.70 31.8 GB 35.2 GB Won't fit — 5.87
IQ3_XXS 3.06 26.6 GB 30.0 GB Won't fit 6K 20.9
Q2_K 2.63 23.1 GB 26.5 GB Fits, but tight 17K 47.5
IQ2_XXS 2.06 18.4 GB 21.8 GB Runs comfortably 32K 58.5
IQ1_M 1.75 15.9 GB 19.3 GB Runs comfortably 40K 66.8

Also worth checking

Qwen2.5 72B on every GPU
Full hardware tableEverything that runs on a RTX 5090
Full model tableWhy the KV cache is the part that bites
Guide