On November 11 the official Qwen account announced "Qwen4 is coming" and opened a preview of its flagship Qwen4-Max, pitching deeper knowledge, richer world understanding, and stronger coding and agentic capabilities. According to the official QwenCloud docs, the model ID is qwen4-max-preview, with a 1M-token context and up to 393,216 output tokens shared by thinking and the final answer. Thinking is on by default, and the model accepts image and video input, supports function calling, web search, and structured outputs, and is compatible with the OpenAI Chat Completions, Responses, and Anthropic Messages APIs. During the preview it is available only through a Token Plan subscription, with no pay-as-you-go access; QwenWork, Qoder, the Qwen app, and QwenStudio also offer early access. No benchmarks or open-weight plans have been published.
Key Takeaways
- ✓Specs: 1M-token context and up to 393,216 output tokens, with thinking and the final answer sharing one budget (per the official docs)
- ✓Reasoning effort:
reasoning_effortoffers 3 levels,low/medium/xhigh, defaulting toxhigh; do not set it together withthinking_budget - ✓Access: Token Plan only, no pay-as-you-go; base URL
https://token-plan.maas.qwencloudapi.com, compatible with both OpenAI and Anthropic protocols (2 protocol families) - ✓Plan quotas: Token Plan Individual gives 20,000 (Lite) to 200,000 (Pro) Credits a month; Lite/Essential normally start at $6/month, with new users from $3 during the current promotion (official figures)
- ✓Migration note: starting with
qwen4-max-preview,max_tokensin Chat Completions equalsmax_completion_tokens, so a limit sized only for the answer may get truncated by thinking; no benchmarks have been published
Key Decision Metrics at a Glance
Turn your technical choice into a development budget
Compare 40 dev plans & simulate token costs vs $20/mo subscriptions
Project Links & Resources
Direct AccessIn-Depth Technical Analysis
Background and Industry Pain Points
Qwen is one of the most active Chinese model families across both open weights and APIs, and its Max line has long been the closed flagship. The official account teased "The becoming begins" on November 6, and about a week later, on November 11, it announced that Qwen4 is coming and opened a preview of Qwen4-Max first. The target use cases it highlights are complex coding, knowledge work, front-end builds, and professional software workflows, which puts it straight into coding agents, currently the most contested segment.
Architecture Highlights and Underlying Mechanisms
The official docs describe Qwen4-Max-Preview as the next-generation flagship multimodal model, combining text generation, image and video understanding, and deep reasoning in one model. Thinking is on by default and can be tuned with reasoning_effort across low, medium, and xhigh. Multi-turn chats enable preserve_thinking by default, so prior thinking should be returned in the reasoning_content field. Context is 1M tokens, with up to 393,216 output tokens. Parameter count, architecture details, and training data have not been disclosed.
Authoritative Benchmarks and Measured Comparisons
As of publication, Qwen has released no benchmarks or technical report and no comparison with Claude, GPT, or Gemini, so we do not speculate here. The launch post simply asks developers to test it on a real problem. Capabilities will keep changing during the preview, and the model will be retired or replaced by an official version afterward, so early community results may not hold across versions.
Developer Adoption and Getting Started
Subscribe to Token Plan Individual or Team, get a dedicated key from the API Keys page, set the base URL to https://token-plan.maas.qwencloudapi.com, and use the model ID qwen4-max-preview. Token Plan officially supports tools such as Qwen Code, Claude Code, Cursor, Cline, OpenCode, and Codex. Note that max_tokens now includes thinking tokens, so raise old limits to avoid truncated answers. If you would rather not write code, try it early in the Qwen app, QwenWork, Qoder, or QwenStudio.
Benchmark side-by-side against alternatives, or calculate monthly token cost vs subscription break-even.
Discussion & Comments
0Sign in to join the discussion
Connect with AI developers to exchange benchmark insights.