Alibaba has unveiled Qwen 3.8, its latest open-weight AI model featuring 24 trillion parameters, claiming performance on par with leading frontier AI models and trailing only Anthropic’s Claude Fable 5 in select benchmarks. A preview is available through Alibaba Token, Qoder, and Qoder Work, with the latter two reportedly priced at just 10% of the standard cost. Alibaba states Qwen 3.8 significantly outperforms Qwen 3.7 Max in coding, reasoning and complex productivity tasks. The launch appears aimed at slowing the momentum of Moonshot AI’s Kimi K3, a 2.8-trillion-parameter model unveiled days earlier. Moonshot said Kimi K3 delivers frontier-level performance across benchmarks and outperformed Anthropic’s Claude Fable 5 and OpenAI’s GPT-5.6 SOL on ProgramBench and SWE Marathon, based on self-reported data. Moonshot, which crossed $300 million in annual recurring revenue in June, reportedly prepares for an IPO within the next six months. Alibaba revealed Qwen 3.8 will soon be released as an open-weight model.