Alibaba launches Qwen4-Max and claims lead over rivals in coding benchmark
Alibaba has unveiled Qwen4-Max, the flagship of its new Qwen4 series, and says the model beats Claude Opus 5.5 on coding tasks. The company is pitching the model as much on efficiency and cost as on raw performance.
Benchmark results and their limits
According to the research findings on the launch, Qwen4-Max scored 75.7 on the DeepSWE v1.1 coding benchmark. That result placed it ahead of Claude Opus 5.5 and showed competitive results against Kimi K3.1 Pro.
The same material places Qwen4-Max behind competitors such as Kimi on some measures, even while highlighting its favourable cost per task. The benchmark claims come from the company and the report describing its launch.
Architecture and capabilities
The model runs on a sparse mixture-of-experts architecture built around a 2.4-trillion-parameter backbone. Rather than activating every parameter for each request, such a design routes each task to a subset of internal experts, so a very large model need not incur the full computing cost on every job.
Qwen4-Max is multimodal, handling text, images and video, and supports a context window of up to 1 million tokens, which allows it to work with long documents or large codebases in one session.
Efficiency and pricing
Alibaba's central argument is efficiency. Qwen4-Max uses about 50% fewer tokens than its predecessor, Qwen3.8-Max, and average task costs fell by more than 60% compared with that model. Because AI providers typically bill by token, fewer tokens per job means a smaller invoice for the same work.
Pricing is set at $1.50 per million input tokens and $4.50 per million output tokens.
Access for developers
Developers can reach Qwen4-Max through an API on QwenCloud. Alibaba has also made the model compatible with protocols established by OpenAI and Anthropic, so teams already working in those ecosystems can switch with a configuration change rather than a rewrite.
Open questions
- Will independent testing confirm the coding benchmark results?
- How will rivals respond on price and efficiency?