Rent a GPU

Can the NVIDIA RTX 3090 run Qwen2.5 14B?

RTX 3090 × Qwen2.5 14B8K context
YesRuns fully on the GPU
Memory at Q8_017.8 GB of 24 GB
Weights 15.6 GBKV cache 1.6 GBRuntime 0.6 GB
QuantQ8_0Near-lossless
Memory17.8 GBof 24 GB usable
Speed~36tokens 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

Yes. The NVIDIA RTX 3090 can run Qwen2.5 14B fully in VRAM at Q8_0 (near-lossless), using about 17.8 GB of the 24 GB available, at roughly 36 tokens/second.

Try other context lengths

Run it with Ollama

With Ollama installed, one command downloads and starts Qwen2.5 14B:

ollama run qwen2.5:14b

Ollama's default download is Q4_K_M (about 11.1 GB here). You have room for Q8_0, which is near-lossless: pick that GGUF in LM Studio or from Hugging Face.

New to this? Step-by-step setup for the RTX 3090

Run it in the cloud

No NVIDIA RTX 3090? Run Qwen2.5 14B in the cloud.

Rent a cloud GPU by the hour. No upfront cost, stop any time.

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 24 GB the NVIDIA RTX 3090 can use.

QuantQualityMemoryFits?Speed
FP16Full quality31.6 GBToo big—
Q8_0Near-lossless17.8 GBFits~36 tok/s
Q6_KExcellent14.3 GBFits~47 tok/s
Q5_K_MVery good12.7 GBFits~54 tok/s
Q4_K_MGood, most popular11.1 GBFits~63 tok/s
Q3_K_MNoticeably degraded9.4 GBFits~78 tok/s
Q2_KHeavily degraded7.7 GBFits~102 tok/s

GPUs that run Qwen2.5 14B comfortably

Models that run well on the NVIDIA RTX 3090

Frequently asked questions

Can NVIDIA RTX 3090 run Qwen2.5 14B?

Yes. The NVIDIA RTX 3090 can run Qwen2.5 14B fully in VRAM at Q8_0 (near-lossless), using about 17.8 GB of the 24 GB available, at roughly 36 tokens/second.

How much VRAM does Qwen2.5 14B need?

About 11.1 GB at Q4_K_M with 8,192 tokens of context, and about 31.6 GB at full FP16 precision.

How fast is Qwen2.5 14B on the NVIDIA RTX 3090?

Roughly 36 tokens per second at Q8_0, estimated from the NVIDIA RTX 3090's 936 GB/s memory bandwidth.