SyncAI.news, a Varaisys broadcasting
Alibaba launches Qwen4-Max AI model, claims it beats GPT-6.1-Sol and Claude Opus 5.5 in coding
MA

Mint: AI

· 1 min read

IndiaMint: AI

Alibaba launches Qwen4-Max AI model, claims it beats GPT-6.1-Sol and Claude Opus 5.5 in coding

Alibaba has launched Qwen4-Max, its latest flagship AI model, claiming it outperforms OpenAI's GPT-6.1-Sol and Anthropic's Opus 5.5 in coding benchmarks. The model features 2.4 trillion parameters, advanced cybersecurity safeguards and a claimed 62% reduction in average AI task costs.

Alibaba Group on Monday launched Qwen4-Max, its latest flagship artificial intelligence (AI) model, as the Chinese technology company looks to strengthen its position in the global AI race against US rivals OpenAI and Anthropic. The new model, according to the company, focuses on coding, cybersecurity, scientific research and complex professional tasks. Alibaba says that its Qwen4-Max can match or surpass top US AI models on some benchmarks and reduce task execution time.

The new model is meant to be used for more heavy-duty tasks, such as writing and debugging code, analysing financial documents, building games and assisting with scientific experiments. Alibaba has also introduced additional safeguards to restrict potentially harmful uses, such as in cybersecurity and chemical, biological, radiological and nuclear (CBRN) applications. The launch follows growing concerns about the threats posed by more sophisticated AI systems.

Alibaba Qwen4-Max specifications and features

Alibaba said Qwen4-Max is based on its latest Qwen4 architecture and is equipped with 2.4 trillion parameters, of which 71 billion are active per token. It employs a Mixture-of-Experts (MoE) design, with only the necessary portions of the model engaged in particular tasks.

It also has a context window of up to one million tokens and native vision capabilities to process visual information alongside text.

According to Alibaba, Qwen4-Max is 35% more efficient in terms of tokens on its Work360 benchmark and 49% more efficient on Terminal-Bench 4.0 than its predecessor, Qwen3.8-Max. The company also saw a 62% drop in the average cost of its tasks on Work360, decreasing costs from $4.43 to $1.67.

Pragya Singha Roy

Original source

This story was published by Mint: AI. SyncAI.news shows a preview; the complete article is on the publisher's site.

Read the full story on livemint.com

Similar News