Qwen2.5 72B Instruct at Q4_K_M needs about 46.6 GB, more than the 12 GB of the RTX 4070 SUPER: part of the model runs from system RAM, which slows generation. At Q3_K_M it does not fit in VRAM.
Q4_K_M · 8K context
The model needs about 46.6 GB but only 11.6 GB of VRAM is usable: 34.9 GB must be loaded into system RAM. It runs, 25% on the GPU, but more slowly.
Estimated speed is at most 1.1 tokens/s: too slow to be usable for chat.
~46.6 GB (45.3 GB–47.9 GB) needed · estimate
34.9 GB must be offloaded to system RAM
EstimateData updated 24 days agoSource: Run LLM Locally engine, Hugging Face config.json + model card
What makes this estimate less reliable
| Qwen2.5 72B Instruct Q8_08K~79.3 GB · No data | Incompatible |
|---|---|
| Qwen2.5 72B Instruct Q6_K8K~62.1 GB · 0.4–0.8 tok/s estimate | Not recommended |
| Qwen2.5 72B Instruct Q5_K_M8K~54.1 GB · 0.5–0.9 tok/s estimate | Not recommended |
| Qwen2.5 72B Instruct Q4_K_M8K~46.6 GB · 0.6–1.1 tok/s estimate | Not recommended |
| Qwen2.5 72B Instruct IQ4_XS8K~42.4 GB · 0.6–1.3 tok/s estimate | Not recommended |
| Qwen2.5 72B Instruct Q3_K_M8K~38.2 GB · 0.7–1.5 tok/s estimate | Not recommended |
Prices and links follow the market you choose.
AMD
CompatibleRuns Qwen2.5 72B Instruct Q4_K_M entirely in memory
Unified memory PC
CompatibleRuns Qwen2.5 72B Instruct Q4_K_M entirely in memory
Lowest memory tier
ExcellentRuns Qwen2.5 72B Instruct Q4_K_M entirely in memory
No current offer imported for this market: the link opens a marketplace search.
