Meituan releases 1.6‑trillion‑parameter agent model built without Nvidia GPUs
Original headline: Meituan releases a 1.6-trillion-parameter agent model that never touched an Nvidia GPU
Meituan has released the weights for LongCat-2.5, a 1.6‑trillion‑parameter AI agent model that was trained entirely on Chinese hardware, with no Nvidia GPUs involved. The model scored 88.4 on the Terminal Bench, just behind Claude Opus 5, and was demonstrated using 194 Huawei Ascend chips.
The achievement demonstrates that ultra‑large AI models can be built on domestically sourced hardware, underscoring a path toward technological self‑reliance.
Research angleHow does building ultra‑large AI models on in‑house hardware affect a firm's AI competitiveness? — Resource‑Based View (RBV)
This story could become your next paper
HJITM works with researchers from study design through surveys, analysis and submission to international journals in generative AI and technology management. Bring us your research question.
This item is an AI-generated summary of reporting by The Decoder. Please check the original article for accuracy. Copyright in the original article belongs to The Decoder.