Alibaba has unveiled an AI model comparable to Anthropic's cutting-edge developments

China’s Alibaba has unveiled Qwen3.8-Max, the largest AI model in its history. According to the company, the new model is comparable in performance to Anthropic’s cutting-edge developments and, in some tests, outperforms them. Against this backdrop, Alibaba’s shares in Hong Kong rose 7.3% in a matter of moments, marking their biggest gain in nearly a month, Bloomberg reports.

Qwen3.8-Max is built on 2.4 trillion parameters (the more parameters an AI model has, the greater its capabilities). According to Alibaba, based on the results of some tests, it outperforms the Kimi K3 model from the Chinese AI startup Moonshot on several benchmarks, and also shows comparable or—in some cases—better results than Anthropic’s Fable 5. In July, U.S. authorities temporarily imposed export restrictions on this cutting-edge AI model due to its extensive capabilities.

The new model performs well at autonomous code generation and long-term tasks. According to Alibaba, during internal testing, it was able to independently complete a software development project over the course of 16 days. In addition, the system operates more efficiently than its predecessors: when performing tasks, it activates only a portion of its components, which reduces computational costs and response latency, notes Bloomberg.

Qwen3.8-Max will be available starting next week—users will be able to download and adapt the technology to their specific tasks. The usage fee is set at $2 per 1 million input tokens and $6 per 1 million output tokens. According to a Bloomberg analysis, this is several times cheaper than some U.S. alternatives. For example, using Anthropic’s cutting-edge Fable 5 costs $10 per 1 million input tokens and $50 per 1 million output tokens. Comparable rates are set for using OpenAI’s most powerful AI model, Sol—$5 and $30, respectively.

This article is being updated

This article was AI-translated and verified by a human editor