Overview
Qwen3.8-Flash is a fast multimodal Qwen model for reasoning and generation. Alibaba announced it on August 26, 2026. It is intended for coding assistance, agentic workflows and visual understanding, including code repair, desktop application operation, and analysis of charts and long videos. The service is compatible with OpenAI and Anthropic API protocols. (docs.qwencloud.com)
Capabilities and uses
- Accepts text, images and video; produces text. Supports function calling, structured output and context caching. (docs.qwencloud.com)
- Suitable for processing long documents, codebases and complex conversations, as well as building high-concurrency applications and intelligent workflows. (docs.qwencloud.com)
Limits
- The context window is 1,000,000 tokens. Maximum input is 991,808 tokens, or 983,616 in thinking mode; maximum output is 131,072 tokens. (docs.qwencloud.com)
- Features vary by deployment: Beijing supports web search and batch inference; Singapore's International scope supports web search but not batch inference. The listed Frankfurt, Tokyo, Virginia and Hong Kong scopes support neither. Fine-tuning is not supported in the listed deployments. (help.aliyun.com)
- Outputs should be checked for accuracy, particularly before consequential use.
Official API pricing
The following are original per-token invocation prices, excluding limited-time promotions. Prices are in the currency published by each platform and are not conversions. Queried November 16, 2026.
China mainland: Alibaba Cloud Model Studio
For China (Beijing), per 1 million tokens: input CNY 0.8, output CNY 2.7, input cache hit CNY 0.1, explicit cache creation CNY 1.25, and explicit cache hit CNY 0.1. Official domestic pricing source. (help.aliyun.com)
International: QwenCloud
For Singapore, International scope, per 1 million tokens: input USD 0.15, output USD 0.47, implicit-cache input USD 0.016, explicit cache creation USD 0.20, and explicit cache read USD 0.016. The same listed rates apply to Frankfurt EU, Hong Kong scope, and Virginia US scope. Official international pricing source. (docs.qwencloud.com)
For Global scope in Frankfurt, Hong Kong, Tokyo or Virginia, per 1 million tokens: input USD 0.113, output USD 0.382, implicit-cache input USD 0.014, explicit cache creation USD 0.177, and explicit cache read USD 0.014. These are separate deployment-scope prices, not exchange-rate equivalents of the Beijing CNY prices. (docs.qwencloud.com)