Alibaba
Qwen4-Max
multimodal reasoning with a million-token context

Key facts
- 16 Nov 2026production model
- Released
- 1,000,000tokens
- Context
- $1.50 / $4.50per million, Singapore international
- Input / output
- 262,144reasoning plus answer tokens
- Thinking output
Qwen4-Max is Alibaba’s production AI model for text, images and video. It supports reasoning, coding, tools and structured responses, with region-specific API prices.
Qwen4-Max is Alibaba’s production AI model for understanding text, images and video and returning text. It can reason through a task, write code, use tools and produce structured responses. QwenCloud lists its production release on 16 November 2026.
Production access is separate from the preview
The model ID is qwen4-max. The production release supports pay-as-you-go billing as well as Token Plan access. Its model documentation sets out supported regions and request limits.
The earlier Qwen4-Max-Preview remains a separate, time-limited release that retires after 25 November in UTC+8. Its larger preview output allowance should not be copied into production specifications.
Input and reasoning have separate limits
The context window is 1,000,000 tokens. Maximum input is 991,808 tokens without thinking or 983,616 with thinking. Output is capped at 131,072 tokens without thinking; with thinking, 262,144 tokens cover both reasoning and the final answer. Tokens are pieces of encoded content, rather than words.
These are individual limits in the provider’s specification. A request must also fit the overall context budget; the largest input and output limits cannot automatically be added together.
Prices depend on the deployment region
The pricing table lists Singapore international rates of $1.50 input, $4.50 output and $0.075 cached input, per million tokens. Beijing/global rates are separately listed at $1.413 input and $4.24 output, with their own cached-input terms.
Choose the intended region before comparing costs. Standard usage, cached requests and subscription allowances are different billing arrangements. Evaluate the model on representative tasks and verify its answers and tool actions before relying on them.
More in Large Language Models
All LLMs →- Google DeepMindGemini 4 Argonlonger reasoning, coding and cyber defence
- OpenAIGPT-5.6the flagship from 9 July to 3 September 2026
- OpenAIGPT-6 Astrathe new generation, rolling out from 3 September 2026
- AnthropicClaude Fable 5.2An AI model for difficult coding, research and long-running tasks
- Google DeepMindGemini 4 CarbonCoding, research and tasks with many steps, now available more widely
- OpenAIGPT-6.1 Solthe upgraded mid-price GPT-6 model, released 29 September 2026