Rent a GPU

Can the NVIDIA RTX 4080 Super run Qwen2.5 72B?

RTX 4080 Super × Qwen2.5 72B8K context
NoDoes not fit in memory
Memory at Q4_K_M47.4 GB of 16 GB
Weights 44.1 GBKV cache 2.7 GBRuntime 0.6 GB
On the GPU29%of the Q4_K_M weights
In system RAM31.4 GBthe rest spills over
Offload speed~1.1tokens per second
  1. FP16
  2. Q8_0
  3. Q6_K
  4. Q5_K_M
  5. Q4_K_M
  6. Q3_K_M
  7. Q2_K

No. Even at Q4_K_M, Qwen2.5 72B needs about 47.4 GB, but the NVIDIA RTX 4080 Super has 16 GB usable. You could offload some layers to system RAM, but it will be very slow.

Try other context lengths

Run it on a rented GPU

On a cloud GPU with enough memory, the same command you'd use locally works:

ollama run qwen2.5:72b

At Q4_K_M it needs about 47.4 GB, so pick a GPU with at least 48 GB.

New to this? Step-by-step setup for the RTX 4080 Super

Run it in the cloud

Too big for the NVIDIA RTX 4080 Super? Run Qwen2.5 72B on a rented GPU instead.

Rent a cloud GPU by the hour. No upfront cost, stop any time. Not sure whether to rent or buy? Compare the cost.

Rent a GPU on RunPodRent a GPU on Vast.ai

Memory needed by quantization

At 8,192 tokens of context. The black line marks the 16 GB the NVIDIA RTX 4080 Super can use.

QuantQualityMemoryFits?Speed
FP16Full quality148.7 GBToo big—
Q8_0Near-lossless80.5 GBToo big—
Q6_KExcellent62.9 GBToo big—
Q5_K_MVery good55 GBToo big—
Q4_K_MGood, most popular47.4 GBToo big—
Q3_K_MNoticeably degraded38.8 GBToo big—
Q2_KHeavily degraded30.5 GBToo big—

Measured speed from real RTX 4080 Super owners

Tokens per second that people measured on their own NVIDIA RTX 4080 Super, next to our estimate.

No one has measured Qwen2.5 72B on the NVIDIA RTX 4080 Super yet. The speeds above are estimates. Run the benchmark and yours becomes the first real number on this page.

Have an RTX 4080 Super? Add your real speed

One command times Qwen2.5 72B in your own Ollama, shows you exactly what it would send, and asks before sending anything. No sign-up, no personal data.

macOS and Linux

curl -fsSL https://gpufit.dev/bench.sh | sh -s -- qwen2.5:72b

Windows (PowerShell)

$env:GPUFIT_MODEL="qwen2.5:72b"; irm https://gpufit.dev/bench.ps1 | iex

How the benchmark works

GPUs that run Qwen2.5 72B comfortably

Check Mac Studio M4 Max (64 GB) prices

Models that run well on the NVIDIA RTX 4080 Super

Frequently asked questions

Can NVIDIA RTX 4080 Super run Qwen2.5 72B?

No. Even at Q4_K_M, Qwen2.5 72B needs about 47.4 GB, but the NVIDIA RTX 4080 Super has 16 GB usable. You could offload some layers to system RAM, but it will be very slow.

How much VRAM does Qwen2.5 72B need?

About 47.4 GB at Q4_K_M with 8,192 tokens of context, and about 148.7 GB at full FP16 precision.

How fast is Qwen2.5 72B on the NVIDIA RTX 4080 Super?

It does not fit in the NVIDIA RTX 4080 Super's memory, so most layers would run from system RAM and generation would be very slow.