Meituan's LongCat team releases 1.6T-parameter AI model trained on domestic ASICs

LongCat-2.0 is billed as an open-source MoE model, but the release still says full weights are coming soon.

By · Published

Primary source: LongCat

Why it matters

LongCat-2.0 turns Meituan's AI work into a test of whether frontier-scale training can move off Nvidia-dependent infrastructure without falling out of developer workflows.

Illustration of a giant cat silhouette formed from ASIC chips and circuit traces representing Meituan's LongCat-2.0, a 1.6T-parameter model trained on domestic ASICs.

Meituan LongCat (@Meituan_LongCat) released LongCat-2.0 on June 30th, positioning the 1.6 trillion-parameter mixture-of-experts model as proof that a Chinese internet company can train and serve a frontier-scale system on domestic AI ASIC infrastructure rather than the Nvidia stack that still defines most large-model economics.…

Reader comments

Conversation for this story loads after sign-in.