Qwen3.5-397B-A17B
Legacy · availableA Qwen model with a 256K-token context window, image input and tool calling. On the Alibaba Cloud Model Studio API since February 15, 2026.
Access details: Alibaba Cloud Model Studio API
The model ID to use in API requests.
qwen3.5-397b-a17bOpen weights on Hugging Face: Qwen/Qwen3.5-397B-A17B
Capabilities & limits
- Context window
- 256K
- Maximum output
- 64K
- Image input
- Supported
- Tool calling
- Supported
- Structured output
- Supported
Pricing
USD / 1M tokens- Input
- $0.60
- Output
- $3.60
Standard rates per million tokens. Discounts and long-context rates, where offered, are on Qwen's pricing page.
Full rate card 3 rates
All prices in USD.
Input tokens
$0.60 per 1M tokens- input 0 to 256K tokens per request
Output tokens, non-thinking mode
$3.60 per 1M tokens- input 0 to 256K tokens per request
- non-thinking mode
Output tokens, thinking mode
$3.60 per 1M tokens- input 0 to 256K tokens per request
- thinking mode (chain of thought plus answer)
USD, International deployment scope (Singapore region) on the international pricing page. Standard prices; the page notes limited-time promotions are excluded. Context-cache and explicit-cache rates are described on a separate page and were not captured; where the page flags a 50% batch discount it is noted here. Rates for other regions (China, Germany, US, Japan, Hong Kong) are not captured. Text-generation page says 256k context but the vision-model page says 32k, so no context is recorded.
Price history
Input $0.60 per million tokens
Output $3.60 per million tokens
Technical details
- Parameters
- 397B total
- Active parameters
- 17B
- Architecture
- Hybrid linear attention (Gated Delta Networks) with sparse mixture of experts
- Input modalities
- Text, image, video
- Output modalities
- Text
From Qwen
Official announcement ↗How they describe it
the open-weight of the first model in the Qwen3.5 series
Release & lifecycle
- First available
- Feb 15, 2026
- legacy
- As of Oct 1, 2026