Zhipu sharpens China’s AI edge in race with US giants

Chinese AI company Z.ai, also known as Zhipu, has released an updated version of its flagship model, aiming to close the gap with US leaders OpenAI and Anthropic. The new GLM-5.3 model, based on a roughly 700-billion-parameter foundation, focuses on enhancing coding capabilities and offers open access to developers. Z.ai plans to release the model's weights within two weeks, allowing developers to modify and customize the AI system.

Benchmark results indicate that GLM-5.3 significantly outperforms its predecessor and approaches, in some tests, the capabilities of Anthropic's Fable 5. This move comes as Chinese AI companies intensify their efforts to challenge US developers in the rapidly advancing field of large language models. Other Chinese AI companies, such as Moonshot AI's Kimi, DeepSeek's V4 models, and Alibaba Group Holding's Qwen, have also demonstrated strong performance in AI benchmarks.

Z.ai seeks to capitalize on this momentum by offering a permissive license to attract developers and users to its models. The competitive landscape is also shifting on pricing, with DeepSeek increasing the prices of its V4 Flash and Pro models by up to four times. However, after the price increase, the blended price of V4 Pro remains lower than GLM-5.2, according to Artificial Analysis.