Alibaba opens Qwen4-Max preview with 1M context, but no official benchmarks yet

Alibaba's next flagship is in developers' hands for a two-week window. Qwen4-Max preview is available through the Token Plan subscription until November 25, with promotional pricing reportedly as low as a tenth of list. Specs include:

- A 1-million-token context window.

- Up to 393,216 maximum output tokens shared between reasoning and final generation.

- Native tool execution and structured schema output.

- Compatibility with both OpenAI and Anthropic API protocols, which lowers switching costs.

The positioning is agentic office work and complex coding, including front-end and UI generation. Dual-protocol compatibility is a pragmatic move that lets teams point existing Claude- or OpenAI-based tooling at Qwen with minimal changes, a direct appeal to cost-sensitive developers. It arrives the same week DeepSeek raised prices sharply, giving Alibaba an opening to capture price-sensitive workloads.

The competitive context is difficult, though. Alibaba gave no official benchmarks, and early community hands-on results are mixed, especially on interactive UIs and games. Developers say they want official numbers before switching. Compared with Gemini 4 Carbon's published evals and Anthropic's documented Fable 5.2 performance, Qwen4-Max is currently selling on price and openness of protocol.

The broader backdrop is US moves to restrict Chinese access to offshore compute, which sent Alibaba shares down 7%. Those moves could constrain how much capacity Alibaba can devote to serving a frontier model globally. Watch for the official benchmark release, final GA pricing after November 25, and whether open weights follow as with earlier Qwen generations.