Meituan has released LongCat-2.0, a 1.6 trillion parameter model, trained entirely on Chinese AI-ASICs without using any NVIDIA chips. This marks the first documented instance of a frontier model's full pre-training being conducted on domestic Chinese hardware, challenging the notion that top-tier AI development is solely reliant on NVIDIA's accelerators. The model boasts a 1 million token context window and reportedly achieved a score of 59.5 on the SWE-Bench Pro benchmark, surpassing GPT-5.5. AI
IMPACT Challenges the necessity of NVIDIA hardware for frontier model training, potentially accelerating independent AI development in regions with export restrictions.
RANK_REASON Frontier-lab model release with system card and benchmark data. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
