Alibaba Releases Qwen3.8-Max AI Model With 2.4 Trillion Parameters

Alibaba Group Holding Ltd. has released Qwen3.8-Max, its largest AI model to date, featuring 2.4 trillion parameters. The company claims the model's performance rivals global leaders, specifically targeting Anthropic's Fable 5 and Claude Opus 5 models. In internal tests, Qwen3.8-Max demonstrated advanced autonomous coding and long-horizon execution, including completing a software engineering project over 16 days and working continuously for 125 hours to rebuild a mathematics research experiment.

To optimize efficiency, the model activates only specific segments of its parameters during use to reduce latency and computational costs. Alibaba reports that the model outperformed most of 526 human teams in a 24-hour online contest and can independently design computer chips and rewrite research papers. The company has set aggressive pricing at $2 per million input tokens and $6 per million output tokens.

Alibaba intends to make the model's weights freely available for public download next week to validate its claims against U.S. competitors. This move follows a broader trend among Chinese developers, including DeepSeek and Moonshot, to close the performance gap with Western models despite constraints on computing resources. Currently, neither Alibaba nor Anthropic has had these specific performance figures verified by independent researchers.