Nine Chinese AI Firms Launch 10 New Models Since June, Matching U.S. Leaders in Key Benchmarks

Since June 2024, nine Chinese AI companies have launched 10 large models. These include MiniMax’s M3 model, which features programming capabilities and multimodal functionalities; Zhipu AI’s GLM-5.2 model, optimized for ultra-long-horizon tasks; and Kimi’s KimiK3, the world’s first open-source model with 3 trillion parameters. In terms of specific technical specifications, Meituan’s LongCat-2.0 model boasts 1.6 trillion parameters and supports a 1-million-token (1M) ultra-long context window along with native multi-step reasoning. Tencent’s HunYuan Hy3 official version adopts a Mixture-of-Experts (MoE) architecture with a total of 295 billion (295B) parameters. Alibaba’s Qwen3.8-Max model has been upgraded for enhanced comprehensive capabilities in programming, office productivity, and multimodal agents. On cost efficiency, DeepSeek’s V4Pro0813 model averages $0.06 per task—significantly lower than Claude Opus 5’s $2.34. Moreover, Chinese technologies are increasingly adopted by global corporations: Ford utilizes CATL’s battery technology, while Volkswagen collaborates with XPeng to develop intelligent electric vehicles. Chinese innovations in artificial intelligence, batteries, and automotive software are now being deployed globally across electric vehicles, battery systems, and consumer electronics markets.