Arrakis / Models / Qwen3.5-397B-A17B

Qwen3.5-397B-A17B GPU requirements

Qwen3.5-397B-A17B has 403B parameters. To serve it, plan for about 968 GB of GPU memory in 16-bit, 484 GB in 8-bit or 242 GB in 4-bit, including about 20% for the KV cache and activations.

It is a mixture of experts: 512 experts, 10 of them active for each token. All of them stay in GPU memory, so plan memory for all 403B parameters.

Deploy this model Check another model

What it takes to serve it

GPU memory to run it for inference, with the smallest GPU setup that holds it and the lowest price on Arrakis right now.

Precision	Memory and the smallest setup
16-bit	968 GB: 4 × B300, from $18.40 per hour for all 4
8-bit	484 GB: 2 × B300, from $9.20 per hour for all 2
4-bit	242 GB: one B300, from $4.60 per GPU-hour

What it takes to fine-tune it

QLoRA trains adapters on a 4-bit frozen model, LoRA on a 16-bit one, and full fine-tuning updates every weight. Figures are before activations, which grow with batch size and sequence length.

Method	Memory and the smallest setup
QLoRA	282 GB: 4 × H100, from $8.28 per hour for all 4
LoRA	968 GB: 4 × B300, from $18.40 per hour for all 4
Full fine-tuning	6,454 GB: more than one 8-GPU machine

About the model

From its card and files on Hugging Face. Memory figures here follow the GPU memory guide.

Parameters: 403B
Active for each token: 17B
Experts: 512
Experts for each token: 10
Context window (tokens): 262,144
Published in: BF16
License: apache-2.0
Released: 2026-02-16
Weights: Qwen/Qwen3.5-397B-A17B

Deploy it on Arrakis

Open Models in the console, choose vLLM, SGLang or llama.cpp, and rent a GPU it fits on. You get an OpenAI-compatible endpoint with its own API key, and root SSH to the machine. Or run it yourself on any machine with enough GPU memory:

vLLM

vllm serve Qwen/Qwen3.5-397B-A17B --tensor-parallel-size 4

Deploy this model How deploying works

Questions

Live prices for every GPU are on the GPU prices page.

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

To serve it, about 968 GB in 16-bit, 484 GB in 8-bit, 242 GB in 4-bit, including about 20% for the KV cache and activations. To fine-tune it, about 282 GB with QLoRA, 968 GB with LoRA and 6,454 GB for full fine-tuning with Adam, before activations.

Can I run Qwen3.5-397B-A17B on one GPU?

Yes, quantized. In 4-bit it needs about 242 GB, which fits on one B300. In 16-bit it needs 4 B300 GPUs.

What does it cost to run Qwen3.5-397B-A17B?

In 16-bit it needs 4 B300 GPUs; in 4-bit, one B300. Each setup on this page shows the lowest price per GPU-hour on Arrakis right now, and you pay by the hour from a prepaid balance.

How do I deploy Qwen3.5-397B-A17B on Arrakis?

Open Models in the Arrakis console, paste Qwen/Qwen3.5-397B-A17B, choose vLLM, SGLang or llama.cpp, and rent a GPU it fits on. You get an OpenAI-compatible endpoint with its own API key, and root SSH to the machine.

More models

Every model, and a check for any other.