China’s Z.ai targets US AI leaders with new flagship model as coding wars heat up

Chinese artificial intelligence unicorn Z.ai has released a new flagship model, the GLM-4.6, with improved coding abilities, hinting at the start-up’s ambition to compete with American industry leaders Anthropic and OpenAI in the key battleground of coding agents.

Beijing-based Z.ai, formerly known as Zhipu, said on Tuesday that GLM-4.6 had “competitive advantages” over leading domestic and international models such as Anthropic’s Claude Sonnet 4, released in May, and DeepSeek-V3.2-Exp, released on Monday.

According to self-reported scores on eight benchmarks that evaluated coding, reasoning and agentic capabilities, GLM-4.6 showed improvements across the board over its previous version, GLM-4.5, released in July.

However, the model still lags behind Anthropic’s Claude Sonnet 4.5, also released on Tuesday, in coding ability. Anthropic has described its latest model as the “best coding model in the world”.

Coding models are one of the most closely watched applications of the fast-evolving AI industry as experts believe they are a crucial step towards artificial general intelligence, a hypothetical AI system that would outperform humans in most tasks.

Other Chinese AI players such as DeepSeek, Alibaba Cloud and Moonshot AI were also developing the coding abilities of their models.