Alibaba Unveils Qwen 3.8 Max, a 2.4T-Parameter Flagship to Rival Kimi K3

Alibaba Unveils Qwen 3.8 Max, a 2.4T-Parameter Flagship to Rival Kimi K3

Alibaba has launched a preview of Qwen 3.8 Max, a 2.4-trillion-parameter flagship model it says is second only to Fable 5, intensifying China’s open-model rivalry with Moonshot.

Alibaba has unveiled a preview of Qwen 3.8 Max, a 2.4-trillion-parameter flagship model it says is second only to Anthropic’s Fable 5 — a direct shot at Moonshot’s freshly released Kimi K3 in China’s intensifying open-model war.

The preview lands days after Moonshot’s Kimi K3 grabbed headlines. Techmeme surfaced Alibaba’s launch of “a 2.4T parameter Qwen3.8 Max,” and Bloomberg reports Alibaba’s Qwen “unveils preview of flagship AI model,” framing it as the company’s bid to keep pace with domestic rivals.

Alibaba is explicit about the competitive frame. The Decoder says Alibaba’s Qwen takes on Kimi K3 with open-weight Qwen 3.8, “says model is second only to Fable 5,” positioning the new release as the strongest open challenger to both Moonshot and closed US frontier models.

What Qwen 3.8 Max brings

Qwen 3.8 Max is the largest model in Alibaba’s Qwen family to date, at roughly 2.4 trillion parameters, and is being shipped as a preview rather than a finished general release. The scale puts it in the same tier as Kimi K3 and DeepSeek’s largest open systems.

Alibaba’s claim that the model is “second only to Fable 5” is a benchmark positioning statement aimed at developers choosing an open-weight backbone. The Decoder notes the comparison is specifically against Anthropic’s Fable 5, the closed frontier model that has set the bar for coding and reasoning this year.

Beyond the flagship, the Qwen ecosystem keeps expanding. MarkTechPost published a workflow for fine-tuning Qwen3 with LoRA using NVIDIA NeMo AutoModel, signalling that Alibaba wants Qwen to be the default substrate for downstream customization, not just a chat model.

Why the Qwen versus Kimi K3 race matters

The Qwen 3.8 versus Kimi K3 clash is the clearest sign yet that China’s open-model competition has moved from lab novelty to a commercial front. Both Alibaba and Moonshot are now shipping trillion-parameter weights that rival or approach closed US systems on key benchmarks.

For enterprises, the practical effect is leverage: open-weight options from Alibaba, Moonshot, and DeepSeek mean teams can self-host frontier-class inference and avoid per-token API pricing. That dynamic is reshaping procurement across Asia and increasingly in Western markets cautious about vendor lock-in.

The rivalry also raises the geopolitical stakes. As Qwen 3.8 Max challenges US AI leaders, US policymakers are paying closer attention to how quickly open Chinese weights are closing the gap — a theme that now runs through everything from export controls to the debate over a public “World Wide Web of AI.”