
Alibaba launches Qwen4-Max, claims it outperforms US AI rivals in coding and cybersecurity - India Today
Chinese tech giant Alibaba has launched Qwen4-Max, its latest flagship AI model claiming superior performance in coding and cybersecurity compared to US rivals.
Alibaba Group has unveiled Qwen4-Max, its newest flagship artificial intelligence model, positioning it as a superior alternative to US competitors like Anthropic and OpenAI. The Chinese tech giant asserts significant improvements in coding, cybersecurity, professional work, and scientific research capabilities.
Technical Specifications
Built on Alibaba's Qwen4 architecture, the model features a 2.4-trillion-parameter backbone with 71 billion parameters activated per token. It employs a Mixture-of-Experts approach, activating only portions of the model for specific tasks to maintain efficiency. Qwen4-Max supports a one-million-token context window and includes native vision capabilities for processing visual information alongside text.
Performance Claims
Alibaba reports Qwen4-Max is more efficient than its predecessor, using up to 35% fewer tokens on Work360 benchmark and 49% fewer on Terminal-Bench 4.0. The company claims a 62% reduction in average spending per task on Work360, from $4.43 to $1.67. In internal testing, the model scored 75.7 on DeepSWE v1.1 coding benchmark, outperforming Anthropic's Opus 5.5 (74.2) and OpenAI's GPT-6.1-Sol (75.2).
Availability and Pricing
Qwen4-Max is accessible through Alibaba's QwenCloud, QwenWork, Qoder and QwenStudio platforms. The company has introduced Live Delivery, a feature allowing early project versions with subsequent improvements based on user feedback. The cached-input API pricing starts at 0.5 yuan (approximately $0.075) per million tokens. Alibaba plans to release smaller open-source models within the Qwen4 family, though hasn't specified their release schedule.