Qwen3 Coder 30B-A3B Instruct at Q4_K_M needs about 19.8 GB, more than the 16 GB of the RX 7900 GRE: 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 19.8 GB but only 15.5 GB of VRAM is usable: 4.3 GB must be loaded into system RAM. It runs, 78% on the GPU, but more slowly.
~19.8 GB (19.1 GB–20.5 GB) needed · estimate
4.3 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
| Qwen3 Coder 30B-A3B Instruct Q8_08K~33.4 GB · 8–16 tok/s estimate | Possible but slow |
|---|---|
| Qwen3 Coder 30B-A3B Instruct Q6_K8K~26.2 GB · 13–25 tok/s estimate | CPU/GPU offload |
| Qwen3 Coder 30B-A3B Instruct Q5_K_M8K~22.9 GB · 16–33 tok/s estimate | CPU/GPU offload |
| Qwen3 Coder 30B-A3B Instruct Q4_K_M8K~19.8 GB · 23–47 tok/s estimate | CPU/GPU offload |
| Qwen3 Coder 30B-A3B Instruct IQ4_XS8K~17.7 GB · 33–68 tok/s estimate | CPU/GPU offload |
| Qwen3 Coder 30B-A3B Instruct Q3_K_M8K~16 GB · 48–105 tok/s estimate | CPU/GPU offload |
Prices and links follow the market you choose.
AMD
CompatibleRuns Qwen3 Coder 30B-A3B Instruct Q4_K_M entirely in memory
Unified memory PC
ExcellentRuns Qwen3 Coder 30B-A3B Instruct Q4_K_M entirely in memory
Lowest memory tier
CompatibleRuns Qwen3 Coder 30B-A3B Instruct Q4_K_M entirely in memory
No current offer imported for this market: the link opens a marketplace search.
