What hardware runs Qwen3.5 397B A17B?
Qwen3.5 397B A17B is a 396.35B-parameter mixture-of-experts model with 17.35B active parameters (apache-2.0 license). The 3.5 flagship. Frontier-adjacent everything; for nearly everyone, this is a rental, not a purchase.
All figures assume an f16 KV cache, a 0.6 GB display reserve on the GPU, and 64 GB of DDR5 system RAM for the offload tiers. Tune these in the calculator.
Minimum VRAM by quant and context
| Quant | File size | 8K | 16K | 32K | 64K | 128K |
|---|---|---|---|---|---|---|
| Q3_K_M | 165.2 GB | 168 GB | 169 GB | 170.9 GB | 174.8 GB | 182.5 GB |
| Q4_K_M | 227.3 GB | 230.1 GB | 231.1 GB | 233 GB | 236.9 GB | 244.6 GB |
| Q5_K_M | 273.5 GB | 276.3 GB | 277.2 GB | 279.2 GB | 283 GB | 290.8 GB |
| Q6_K | 304.2 GB | 306.9 GB | 307.9 GB | 309.8 GB | 313.7 GB | 321.5 GB |
| Q8_0 | 392.6 GB | 395.3 GB | 396.3 GB | 398.2 GB | 402.1 GB | 409.9 GB |
Full-GPU figures: weights + f16 KV cache + overhead.
GPU compatibility
Fits on GPUExpert offloadPartial offloadCPU only
Quant guidance
Our floor for Qwen3.5 397B A17B is Q3_K_M — below that, quality degrades faster than the VRAM savings are worth (large models tolerate low bpw better than small ones, which is why the floor is lower here). Prefer the highest quant that still lands "Fits on GPU" or “Expert offload” at your context length in the table above.
Frequently asked questions
How much VRAM does Qwen3.5 397B A17B need?
At Q3_K_M and 32K context, Qwen3.5 397B A17B needs about 170.9 GB of VRAM to run fully on GPU (weights + KV cache + overhead). As a mixture-of-experts model it can also run with far less VRAM via expert offload, keeping experts in system RAM.
What is the smallest GPU that can run Qwen3.5 397B A17B?
No single consumer GPU in our set runs Qwen3.5 397B A17B well — it's a cloud-rental or multi-GPU model.
What quantization should I use for Qwen3.5 397B A17B?
We recommend Q3_K_M or higher. Q4_K_M weighs 227.3 GB (4.93 bits/weight); going below Q3_K_M costs noticeable quality on a model this size.
How long a context can Qwen3.5 397B A17B handle?
Qwen3.5 397B A17B supports up to 256K tokens. KV cache grows linearly with context.
Can I run Qwen3.5 397B A17B without a GPU?
Not realistically — even at 8K context the weights don't fit in 64 GB of system RAM.