Z.AI’s GLM-5.2 model challenges Anthropic and OpenAI with lower costs

A Chinese AI company just released a model that performs nearly on par with the best offerings from Anthropic and OpenAI, but charges roughly one-tenth the price.

Z.AI, the global-facing brand of Zhipu AI, launched its flagship GLM-5.2 model in mid-June 2026. The open-weight Mixture-of-Experts model packs roughly 744B to 753B parameters, supports a 1M-token context window, and is specifically tuned for long-horizon agentic coding and reasoning tasks. It’s available under an MIT license on Hugging Face, which means anyone can download it, modify it, and deploy it without asking permission.

The price gap is hard to ignore

GLM-5.2’s API pricing lands at approximately $0.95 to $2 per million input tokens and $3 to $6 per million output tokens.

For comparison, leading closed models from the likes of Anthropic and OpenAI charge between $5 and $15 per million input tokens and $30 to $75 per million output tokens. In English: Z.AI is offering roughly 80-90% savings on the same type of work.

On benchmarks, GLM-5.2 ranks as the top open-source and open-weight model across various coding evaluations. It narrowly trails Claude Opus 4.8 in some tests while outperforming previous GPT models. Community reports have compared its quality output favorably against both Opus 4.8 and GPT-5.5.

The model also offers multiple thinking-effort modes, letting developers choose between faster, cheaper responses and deeper, more thorough reasoning depending on the task.

Who is Zhipu AI, and why should you care

Zhipu AI spun out of Tsinghua University in 2019. The company rebranded its international presence as Z.AI, positioning itself as a global competitor rather than a regional player.

The model is accessible through multiple inference platforms including Together.ai and OpenRouter, which means developers don’t need to navigate unfamiliar Chinese infrastructure to use it.

What this means for developers and investors

The immediate impact falls on development teams, particularly smaller ones that have been priced out of using frontier AI models for production workloads. A team spending $10,000 per month on API calls to a closed model could potentially achieve similar results for $1,000 to $2,000 with GLM-5.2.

For the AI industry’s competitive landscape, when an open-weight model achieves comparable quality at a fraction of the cost, the value proposition of paying for closed models starts to rely heavily on factors like reliability, support, safety guarantees, and ecosystem lock-in rather than raw capability.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.