Qwen3.6 35B-A3B at Q4_K_M needs about 23.1 GB, more than the 20 GB of the RX 7900 XT: part of the model runs from system RAM, which slows generation. At Q3_K_M it fits entirely.
Q4_K_M · 8K context
The model needs about 23.1 GB but only 19.4 GB of VRAM is usable: 3.7 GB must be loaded into system RAM. It runs, 84% on the GPU, but more slowly.
~23.1 GB (22.2 GB–23.9 GB) needed · estimate
3.7 GB must be offloaded to system RAM
EstimateData updated 40 days agoSource: Run LLM Locally engine, Hugging Face config.json + model card
What makes this estimate less reliable
| Qwen3.6 35B-A3B Q8_08K~38.3 GB · 11–21 tok/s estimate | CPU/GPU offload |
|---|---|
| Qwen3.6 35B-A3B Q6_K8K~31.5 GB · 16–31 tok/s estimate | CPU/GPU offload |
| Qwen3.6 35B-A3B Q5_K_M8K~26.6 GB · 23–46 tok/s estimate | CPU/GPU offload |
| Qwen3.6 35B-A3B Q4_K_M8K~23.1 GB · 35–73 tok/s estimate | CPU/GPU offload |
| Qwen3.6 35B-A3B IQ4_XS8K~20.5 GB · 56–123 tok/s estimate | CPU/GPU offload |
| Qwen3.6 35B-A3B Q3_K_M8K~18 GB · 100–186 tok/s estimate | Compatible |
Prices and links follow the market you choose.
NVIDIA
ExcellentRuns Qwen3.6 35B-A3B Q4_K_M entirely in memory
AMD
ExcellentRuns Qwen3.6 35B-A3B Q4_K_M entirely in memory
Apple Silicon
CompatibleRuns Qwen3.6 35B-A3B Q4_K_M entirely in memory
Unified memory PC
CompatibleRuns Qwen3.6 35B-A3B Q4_K_M entirely in memory
No current offer imported for this market: the link opens a marketplace search.
