QW

Qwen3.5-397B-A17B

QwenGGUF by unslothUpdated 2026-03-07
Download on Hugging Face
Parameters397B
Context256K
Layers60
Architectureqwen35moe
Smallest quant106.8 GB
Quick answer

Qwen3.5-397B-A17B needs about 245.6 GB of memory at Q4_K_M (a 244.1 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 29 quants, from 106.8 GB (IQ1_M) to 793.0 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 4.0 GB.

Qwen3.5-397B-A17B: 397B parameter multimodal LLM with 17B active experts

Qwen3.5-397B-A17B is a 397 billion‑parameter causal language model with vision encoder. It uses a qwen35moe architecture and a 262,144‑token context window. The model is licensed under Apache‑2.0 and supports multimodal tasks.

Good at

  • Vision‑language reasoning
  • Long‑context text generation
  • Multilingual support

Limits

  • Requires large VRAM for full precision
  • No built‑in vision tools in GGUF
  • Limited to 262k context tokens

Best for

  • Multimodal research prototypes
  • Large‑scale text generation
  • Vision‑enabled chatbots

Who it’s for. Developers and researchers needing a high‑capacity multimodal model for experimentation. The model is suitable for local inference on high‑end GPUs with sufficient VRAM.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
IQ1_M106.8 GB (4 parts) 108.3 GB
Too big Get file →
IQ2_XXS114.9 GB (4 parts) 116.4 GB
Too big Get file →
IQ2_M123.1 GB (4 parts) 124.6 GB
Too big Get file →
IQ3_XXS140.3 GB (4 parts) 141.8 GB
Too big Get file →
IQ3_S146.4 GB (4 parts) 147.9 GB
Too big Get file →
Q3_K_S164.3 GB (5 parts) 165.8 GB
Too big Get file →
Q3_K_S164.3 GB (5 parts) 165.8 GB
Too big Get file →
Q3_K_M177.4 GB (5 parts) 178.9 GB
Too big Get file →
Q3_K_M177.4 GB (5 parts) 178.9 GB
Too big Get file →
Q3_K_XL178.7 GB (5 parts) 180.2 GB
Too big Get file →
IQ4_XS189.7 GB (5 parts) 191.2 GB
Too big Get file →
IQ4_NL193.8 GB (5 parts) 195.3 GB
Too big Get file →
Q4_K_S228.0 GB (6 parts) 229.5 GB
Too big Get file →
Q4_K_S228.0 GB (6 parts) 229.5 GB
Too big Get file →
MXFP4_MOE237.4 GB (6 parts) 238.9 GB
Too big Get file →
Q4_K_M244.1 GB (6 parts) 245.6 GB
Too big Get file →
Q4_K_M244.1 GB (6 parts) 245.6 GB
Too big Get file →
Q4_K_XL245.3 GB (6 parts) 246.8 GB
Too big Get file →
Q5_K_S276.6 GB (7 parts) 278.1 GB
Too big Get file →
Q5_K_S276.6 GB (7 parts) 278.1 GB
Too big Get file →
Q5_K_M293.7 GB (8 parts) 295.2 GB
Too big Get file →
Q5_K_M293.7 GB (8 parts) 295.2 GB
Too big Get file →
Q5_K_XL294.9 GB (8 parts) 296.4 GB
Too big Get file →
Q6_K326.6 GB (8 parts) 328.1 GB
Too big Get file →
Q6_K326.6 GB (8 parts) 328.1 GB
Too big Get file →
Q6_K_XL362.3 GB (9 parts) 363.8 GB
Too big Get file →
Q8_0421.5 GB (10 parts) 423.0 GB
Too big Get file →
Q8_K_XL427.7 GB (10 parts) 429.2 GB
Too big Get file →
BF16793.0 GB (17 parts) 794.5 GB
Too big Get file →

For 8 GB VRAM, use IQ1_M (106 GB file, ~108 GB RAM). For 16 GB VRAM, IQ2_M (123 GB file, ~125 GB RAM). For 24 GB VRAM, Q4_K_M (244 GB file, ~246 GB RAM). Estimates are approximate.

What GPU do you need for Qwen3.5-397B-A17B?

These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.

  • 4 GB VRAM: no quant fits at an 8K context.
  • 6 GB VRAM: no quant fits at an 8K context.
  • 8 GB VRAM: no quant fits at an 8K context.
  • 12 GB VRAM: no quant fits at an 8K context.
  • 16 GB VRAM: no quant fits at an 8K context.
  • 24 GB VRAM: no quant fits at an 8K context.
  • 32 GB VRAM: no quant fits at an 8K context.
  • 48 GB VRAM: no quant fits at an 8K context.

Which Qwen3.5-397B-A17B quant should you choose?

Q4_K_M is the usual starting point for Qwen3.5-397B-A17B: it balances file size and quality well. Here it is a 244.1 GB file.

The Q8_0 file is 421.5 GB, about 1.7 times the size of Q4_K_M. It keeps the weights closer to the original, which only matters if you have the memory to spare.

If memory is tight, the smallest file is IQ1_M at 106.8 GB, with the quality trade-offs that come with fewer bits. Use it only when memory is very tight. Expect a clear drop in quality, which is usually worse on small models and easier to live with on large ones.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architectureqwen35moe
Parameters397B (approximate, mixture of experts)
Active parameters17B
Layers60
Trained context262,144 tokens
KV cache (fp16)120 KB per token

Source

Quantized byunsloth
FamilyQwen
Licenseapache-2.0
Downloads (HF)19,770
Last updated2026-03-07

Frequently asked questions about Qwen3.5-397B-A17B

How much VRAM does Qwen3.5-397B-A17B need?

About 245.6 GB at Q4_K_M with an 8K context window. That is the 244.1 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.

Can I run Qwen3.5-397B-A17B on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 108.3 GB, so you would need a larger GPU or CPU offloading.

What is the file size of Qwen3.5-397B-A17B GGUF?

It depends on the quant: IQ1_M is 106.8 GB, IQ2_XXS is 114.9 GB, IQ2_M is 123.1 GB, IQ3_XXS is 140.3 GB, IQ3_S is 146.4 GB, Q3_K_S is 164.3 GB and 23 more.

What context length does Qwen3.5-397B-A17B support?

The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 1.0 GB at 8K and 4.0 GB at 32K.

Where can I download Qwen3.5-397B-A17B GGUF?

The files are on Hugging Face in the unsloth/Qwen3.5-397B-A17B-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.

What license does Qwen3.5-397B-A17B use?

Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.

Can I run Qwen3.5-397B-A17B on 8 GB VRAM?

No, the smallest quant requires over 100 GB of VRAM. You need at least 8 GB of VRAM for the smallest quant, but the model is still too large for that budget.

Does the GGUF version support vision?

Yes, the GGUF files include vision encoder support as indicated by detected abilities.

What license applies to this model?

The model is released under the Apache‑2.0 license.

Other GGUFs of Qwen3.5-397B-A17B

Similar models

Browse all →

More from unsloth

All by unsloth →

More Qwen models

All Qwen →

Other large models

All large models →