Alibaba launches Qwen4-Max-Preview with 1 million-token context
The preview model is limited to Token Plan subscribers and will be pulled from service on November 25, with no weights or benchmarks published.
Alibaba announced and released Qwen4-Max-Preview, the first model in its Qwen4 family, on November 11, 2026, according to aitimes.com. The model is being offered as a limited preview through the subscription-based Token Plan API and inside several Alibaba products.
Qwen4-Max-Preview supports a context length of up to 1 million tokens and a maximum output of 393,216 tokens, counting reasoning and the final answer together. Reasoning is switched on by default and its intensity can be adjusted.
The model accepts text, images and video as input. It also supports web search, function calling and built-in tools, JSON-schema-based structured output, and compatibility with OpenAI and Anthropic APIs.
Subscription-only access
Alongside the Token Plan API, Qwen4-Max-Preview is available in QwenWork, Qoder, Qwen APP and QwenStudio. API access is restricted to Token Plan subscribers, and pay-as-you-go usage is not supported.
Alibaba said the preview runs from the 11th to the 25th of the month, after which the model will be removed from service. The company did not publish model weights, benchmark results or a technical report.
Quick answers
How long can Qwen4-Max-Preview's context be?
The model supports a context length of up to 1 million tokens, with a maximum output of 393,216 tokens that includes both reasoning and the final answer.
Where can I use Qwen4-Max-Preview?
It is available through the subscription-based Token Plan API and in QwenWork, Qoder, Qwen APP and QwenStudio. API access is limited to Token Plan subscribers; pay-as-you-go usage is not supported.
How long will the Qwen4-Max-Preview be available?
The preview runs from the 11th to the 25th of the month, after which Alibaba will remove the model from service.