No — not on this device
Qwen3.5 397B-A17B at Q4_K_M needs 227.4 GB against 204.8 GB usable, and the shortfall of 22.6 GB is more than 64 GB of system RAM can cover at a tolerable speed. A smaller sibling or a lower quantisation is the honest answer here.
Flagship-class at 17B active. A 256 GB Mac Studio or a multi-GPU box at Q4.
What hardware do I need for Qwen3.5 397B-A17B? →
Fits instead: Qwen3.5 122B-A10B (71.1 GB) · Qwen3.5 9B (6.3 GB)
The VRAM budget
Quantisation ladder
| Quant | Weights | Total @ 8K | Max context | Tok/s | Quality | Fit |
|---|---|---|---|---|---|---|
| Q8_0 | 398.8 GB | 399.6 GB | — | ~3.4 | −0.1% ppl | 194.8 GB over |
| Q6_K | 307.8 GB | 308.6 GB | — | ~4.4 | −0.4% ppl | 103.8 GB over |
| Q5_K_M | 266.0 GB | 266.8 GB | — | ~5.1 | −0.8% ppl | 62.0 GB over |
| Q4_K_M | 226.6 GB | 227.4 GB | — | ~6.0 | −1.9% ppl | 22.6 GB over |
| Q3_K_M | 183.4 GB | 184.3 GB | 256K | 7.4 | −5.4% ppl | Long context |
| Q2_K | 157.2 GB | 158.0 GB | 256K | 8.6 | −15% ppl | Long context |
Quality is the published perplexity delta against f16 weights. Max context assumes an f16 KV cache; q8_0 roughly doubles it. Only 15 of its 60 blocks keep a per-token KV cache; the rest are linear-attention, Mamba or convolution blocks with a fixed-size state.
How to run it
$ llama-server \
-hf Qwen/Qwen3.5-397B-A17B:Q4_K_M \
-c 8192 -ngl 54
The engine underneath most of the others. Every knob is exposed. More on llama.cpp.