impulse

Qwen3.5 397B A17B

Open weights

The Qwen3.5 series 397B-A17B native vision-language model is built on a hybrid architecture that integrates a linear attention mechanism with a sparse mixture-of-experts model, achieving higher inference efficiency.

API id qwen3.5-397b-a17bReleased 16 Feb 2026 (Feb 2026)
Input / 1M
$0.164
at AIHubMix
Output / 1M
$0.986
at AIHubMix
Context window
262K
tokens
Max output
66K
tokens per reply
Hosts
10
10 listings
Released
Feb 2026
16 Feb 2026

Pricing

List prices at AIHubMix, in US dollars per million tokens.

TierInput / 1MOutput / 1M
Standard$0.164$0.986

What it costs a month

At AIHubMix list prices, 30 days.

  • Chat assistant5,000 requests a day, 2K in / 0.5K out$123
  • RAG over documents2,000 requests a day, 8K in / 0.6K out$114
  • Coding agent1,000 requests a day, 40K in / 2K out$256
  • Overnight batch20,000 requests a day, 4K in / 0.4K out$631
Use your own numbers

Where to run it

10 hosts serve Qwen3.5 397B A17B. The comparison is on blended price, 3 parts input to 1 part output.

HostIn / out per 1M
AIHubMixGateway
qwen3.5-397b-a17b
$0.164 / $0.986
qwen3-5-397b-a17b
$0.172 / $1.03
+5% vs headline
DeepInfraInference
Qwen3.5-397B-A17B
$0.45 / $3.00
+194% vs headline
Qwen3.5-397B-A17B-FP8
$0.60 / $3.60
+265% vs headline
qwen3.5-397b-a17b
$0.60 / $3.60
+265% vs headline
qwen3.5-397b-a17b
$0.60 / $3.60
+265% vs headline
GMI CloudInference
Qwen3.5-397B-A17B
$0.60 / $3.60
+265% vs headline
Qwen3.5-397B-A17B
$0.60 / $3.60
+265% vs headline
Together AIInference
Qwen3.5-397B-A17B
$0.60 / $3.60
+265% vs headline
Novita AIInference
qwen3.5-397b-a17b
$0.60 / $3.60
+265% vs headline

Capabilities

Core

  • Streaming
  • Function calling
  • JSON mode
  • Structured outputs

Multimodal

  • Vision
  • Image input
  • Audio
  • Image generation
  • Video

Advanced

  • Reasoning
  • Tool calling
  • MCP
  • Prompt caching
  • Batch API
  • Embeddings
Takes in
TextImages
Produces
Text
Features
Built-in web search
Open weights
License: Apache-2.0
Runs in about 968.2 GB of GPU memory
Runtimes: Ollama, SGLang, llama.cpp, vLLM

Cheaper alternatives

Current models that cost less and keep most of what this one can do: tools, vision, reasoning, structured output, caching and enough context.

Same model, cheaper host

No host sells it for less than AIHubMix.

Smaller Alibaba Qwen models

No cheaper Alibaba Qwen model keeps its key capabilities.

Other frontier labs

  • Claude Haiku 5.5
    Anthropic · on Anthropic
    $0.10 / $0.50
    Has every key capability
    46% less
  • GPT-6 Luna
    OpenAI · on OpenAI
    $0.10 / $0.50
    Has every key capability
    46% less
  • GPT-4o Mini
    OpenAI · on OpenAI
    $0.15 / $0.60
    Lacks reasoning
    29% less

Open-weight picks

  • Muse Glimmer 30B
    Meta · on EmpirioLabs AI
    $0.20 / $0.80
    Has every key capability
    5% less
  • Mistral Small 2603
    Mistral AI · on Mistral AI
    $0.15 / $0.60
    Has every key capability
    29% less
  • Qwen3.8 27B
    Alibaba Qwen · on EmpirioLabs AI
    $0.17 / $0.50
    Has every key capability
    32% less
  • Nemotron 3.5 Super VL
    NVIDIA · on Kyma API
    $0.081 / $0.472
    Has every key capability
    52% less

Call it

Using AIHubMix's API.

Python
from litellm import completion

response = completion(
    model="aihubmix/qwen3.5-397b-a17b",
    messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)

Quick answers

How much does Qwen3.5 397B A17B cost?

On AIHubMix, Qwen3.5 397B A17B costs $0.164 per million input tokens and $0.986 per million output tokens.

Which APIs offer Qwen3.5 397B A17B?

Qwen3.5 397B A17B runs on 10 APIs: AIHubMix, EmpirioLabs AI, DeepInfra, Tensormesh, Scaleway, StreamLake, GMI Cloud, Nebius AI Studio, Together AI and Novita AI.

What is Qwen3.5 397B A17B's context window?

262K tokens, with up to 66K tokens in one reply.

Is Qwen3.5 397B A17B being retired?

No retirement date is scheduled for it on any API we track.

What are cheaper alternatives to Qwen3.5 397B A17B?

Models that cost less and keep most of what it can do: Claude Haiku 5.5 on Anthropic (46% less), GPT-6 Luna on OpenAI (46% less) and Muse Glimmer 30B on EmpirioLabs AI (5% less). Savings are on a mix of 3 input tokens to 1 output token.

When was Qwen3.5 397B A17B released?

Alibaba Qwen released it on 16 Feb 2026.

What is Qwen3.5 397B A17B's API model id?

Use qwen3.5-397b-a17b with AIHubMix. Other hosts use their own ids: qwen3-5-397b-a17b on EmpirioLabs AI, Qwen3.5-397B-A17B on DeepInfra, Qwen3.5-397B-A17B-FP8 on Tensormesh, Qwen3.5-397B-A17B on GMI Cloud.

Sources

Raw catalog entry for qwen3.5-397b-a17b on AIHubMix
{
  "input_cost_per_token": 1.644e-7,
  "litellm_provider": "aihubmix",
  "max_input_tokens": 262144,
  "max_output_tokens": 65536,
  "mode": "chat",
  "output_cost_per_token": 9.864e-7,
  "source": "https://aihubmix.com/api/v1/models",
  "supports_function_calling": true,
  "supports_reasoning": true,
  "supports_response_schema": true,
  "supports_vision": true,
  "supports_web_search": true
}