Z.ai's GLM-4.6 Debuts with Enhanced Coding, But No 'Air' Model Planned

Beijing, China – Z.ai, formerly known as Zhipu AI, has officially launched its new flagship large language model, GLM-4.6, with a strong focus on advanced coding, reasoning, and agentic capabilities. The release comes amidst public anticipation, with a social media post from user Teortaxes▶️, a self-proclaimed "DeepSeek 推特🐋铁粉," stating, "GLM 4.6 Air any day now," suggesting an imminent release of a lighter "Air" version. However, Z.ai has clarified that there are currently no plans to release a smaller "Air" variant of GLM-4.6.

The GLM-4.6 model, available on platforms like Hugging Face and ModelScope, aims to challenge leading models from companies such as Anthropic and OpenAI. Z.ai stated that the new model demonstrates "competitive advantages" over rivals like Anthropic’s Claude Sonnet 4 and DeepSeek-V3.2-Exp, showing improvements across eight benchmarks compared to its predecessor, GLM-4.5. Despite these gains, the company acknowledges that GLM-4.6 still trails Anthropic's Claude Sonnet 4.5 in specific coding benchmarks.

Z.ai, a spin-off from Tsinghua University founded in 2019, has rapidly grown into a significant player in the AI industry, securing substantial funding from investors including Alibaba and Tencent. The company, which rebranded from Zhipu AI in 2025, offers the GLM-4.6 API at a competitive rate of US$0.6 per million input tokens and US$2 per million output tokens on OpenRouter. This strategic pricing and focus on high-performance models underscore Z.ai's ambition to strengthen its position in the global AI market, even as it navigates regulatory challenges, including being blacklisted by the U.S. Commerce Department in January 2025.