What is Qwen3.5-397B-A17B good for?

Reasoning and coding over text, image and video, at the heavy end of the open Qwen3.5 checkpoints and the backbone of Qwen3.5-Plus. Apache 2.0 weights, on a multi-GPU node.

When is Qwen3.5-397B-A17B not a good fit?

Self-hosting on modest hardware: memory follows the full 397B parameters, not the 17B active. No audio input or image output, and no built-in web search, which the hosted Qwen3.5-Plus adds.

What is the cheapest way to run Qwen3.5-397B-A17B?

Hosted, unless you push serious volume. Alibaba Cloud charges $0.60 in / $3.60 out per 1M tokens. The cheapest rental that fits is 8x RTX A6000 at $3,168 a month, which costs the same as about 3B tokens a month on that API.

Can I self-host Qwen3.5-397B-A17B?

Yes. The weights are Apache 2.0 licensed. At 4-bit it needs about 221 GB of GPU memory, which starts at roughly $3,168 a month on the cheapest rental that fits.

Open weights · Apache 2.0

Qwen3.5-397B-A17B

The largest Qwen3.5 open-weight model, a 397B MoE with 17B active, and the backbone behind Qwen3.5-Plus, aimed at reasoning and coding.

Jump to: Specs Pricing Self-hosting Similar models FAQ More from Alibaba Cloud

Visit website Compare vs other LLMs →

List price via Alibaba Cloud

Input: $0.60 /1M tokens
Output: $3.60 /1M tokens

Cheapest: $0.17 / $1.03 per 1M tokens via EmpirioLabs AI

6 providers

Key Specifications

Context window: 262K tokens
Max output: 66K tokens
Released: Feb 2026
Parameters: 397B, 17B active
Inputs: Text, image, video

Capabilities Show details: Reasoning Thinks before it answers, always on or as a switchable mode. Function calling Connect to external tools, APIs, and systems. Structured output Return responses in structured formats like JSON. Code execution Write and execute code in a sandboxed environment. MCP Connect to services via Model Context Protocol.

Hosted API pricing

Input tokens M Output tokens M

Provider	Input /1M tokens	Output /1M tokens	Cached input /1M	Cost at 10M in + 2M out
EmpirioLabs AI Cheapest	$0.17	$1.03		$3.76	View
DigitalOcean	$0.55	$3.50	$0.11	$12.50	View
Alibaba Cloud Creator	$0.60	$3.60		$13.20	View
Lyceum	$0.60	$3.60		$13.20	View
Novita	$0.60	$3.60		$13.20	View
Together AI	$0.60	$3.60	$0.35	$13.20	View

Heads up: Base-tier, on-demand rates per 1M tokens; cached, batch and long-context tiers excluded. A provider may serve a shorter context or a quantized build than the creator's release. Verify before provisioning. More on how we price.

Estimated cost to self-host

Qwen3.5-397B-A17B needs about 221 GB of GPU memory at 4-bit with 32K context. The cheapest rental that fits is 8× RTX A6000 at about $3,168 a month.

That costs the same as roughly 3B tokens a month on Alibaba Cloud's API. Self-hosting is more expensive below that volume.

Single node ≤ 8

Precision	Memory	Cheapest fit (32K context)(Full 262K context)	Cost /mo	Break-even vs API
4-bitINT4 / FP4	221 GB228 GB	8× RTX A6000	$3,168$3,168	3B tokens /mo3B tokens /mo
8-bitFP8 / INT8	400 GB407 GB	8× A100	$10,138$10,138	9B tokens /mo9B tokens /mo
16-bitFP16 / BF16	797 GB804 GB	4× B300	$22,723$22,723	21B tokens /mo21B tokens /mo

Estimates based on median on-demand rates for Nvidia GPUs. Memory is weights plus KV cache for one request, using FP8 KV cache where supported and FP16/BF16 otherwise. Break-even assumes a 5:1 input-to-output ratio. No guarantee of runtime support, usable performance, or that a matching quantized build exists. How we estimate costs.

Similarly priced models

The models nearest Qwen3.5-397B-A17B by blended rate, each at its own cheapest provider.

Model	Blended /1M	Input /1M	Output /1M	Context	Cutoff	vs Qwen3.5-397B-A17B
Llama 4 Maverick Meta	$0.2833	$0.20	$0.70	1M	Aug 2024	−10%
Qwen3-VL-30B-A3B Instruct Alibaba Cloud	$0.2833	$0.20	$0.70	262K		−10%
Muse Glimmer 30B Meta	$0.30	$0.20	$0.80	131K	Jan 2026	−4%
Qwen2.5-7B Instruct Alibaba Cloud	$0.30	$0.30	$0.30	33K		−4%
Qwen3-235B-A22B FP8 Alibaba Cloud	$0.30	$0.20	$0.80	33K		−4%
Qwen3.5-397B-A17B This model Alibaba Cloud	$0.3133	$0.17	$1.03	262K
MiniMax-M3 MiniMax	$0.3167	$0.20	$0.90	1M		+1%
Qwen3-Next-80B-A3B Instruct Alibaba Cloud	$0.325	$0.15	$1.20	262K		+4%
Qwen3-Next-80B-A3B Thinking Alibaba Cloud	$0.325	$0.15	$1.20	262K		+4%
Qwen-MT-Plus Alibaba Cloud	$0.3333	$0.25	$0.75	16K		+6%
Qwen2.5-VL-72B Instruct Alibaba Cloud	$0.3333	$0.25	$0.75	33K		+6%

Prices are USD per 1M tokens at each model's cheapest listed provider. Blended is the cost of 10M input plus 2M output tokens, spread over the 12M.

Thunder Compute

RTX A6000 from $0.35, A100 from $1.09, per GPU per hour

Billed per minute, no commitment, no minimums

One click into VS Code or Cursor, no SSH setup

See live pricing

Frequently Asked Questions

What is Qwen3.5-397B-A17B good for?: Reasoning and coding over text, image and video, at the heavy end of the open Qwen3.5 checkpoints and the backbone of Qwen3.5-Plus. Apache 2.0 weights, on a multi-GPU node.
When is Qwen3.5-397B-A17B not a good fit?: Self-hosting on modest hardware: memory follows the full 397B parameters, not the 17B active. No audio input or image output, and no built-in web search, which the hosted Qwen3.5-Plus adds.
What is the cheapest way to run Qwen3.5-397B-A17B?: Hosted, unless you push serious volume. Alibaba Cloud charges $0.60 in / $3.60 out per 1M tokens. The cheapest rental that fits is 8x RTX A6000 at $3,168 a month, which costs the same as about 3B tokens a month on that API.
Can I self-host Qwen3.5-397B-A17B?: Yes. The weights are Apache 2.0 licensed. At 4-bit it needs about 221 GB of GPU memory, which starts at roughly $3,168 a month on the cheapest rental that fits.

More from Alibaba Cloud

Model	Context	Input /1M	Output /1M
Qwen3-Next-80B-A3B Instruct Open weights	262K	$0.15	$1.20
Qwen3-Next-80B-A3B Thinking Open weights	262K	$0.15	$1.20
Qwen2.5-7B Instruct Open weights	33K	$0.30	$0.30
Qwen3-235B-A22B FP8 Open weights	33K	$0.20	$0.80
Qwen2.5-VL-72B Instruct Open weights	33K	$0.25	$0.75
Qwen-MT-Plus	16K	$0.25	$0.75
Qwen3-VL-30B-A3B Instruct Open weights	262K	$0.20	$0.70
Qwen3.6-27B Open weights	262K	$0.30	$0.60