The number of users of the Chinese model 'GLM-5.2' API on OpenRouter has surged; what are AI users using it for?



'

GLM-5.2 ' is an AI model released by the Chinese company Z.ai on June 17, 2026 (Japan time). It has outperformed Claude Opus 4.7 in several tests and has shown performance exceeding Claude Fable 5 in some tests. OpenRouter, which provides API services for GLM-5.2, has reported on its performance in the first week.

GLM 5.2 - API Pricing & Benchmarks | OpenRouter
https://openrouter.ai/z-ai/glm-5.2#benchmarks




Z.ai is a company known for developing AI models called the GLM series. The GLM series is characterized by its development as an open model, with the models being made publicly available. When the services for 'Claude Fable 5' and 'Claude Mythos 5' were discontinued on June 13, 2026, Z.ai announced the release of GLM-5.2, which was officially released on June 17, 2026. GLM-5.2 has achieved scores that surpass Claude Opus 4.7 and GPT-5.5 in multiple benchmark tests, and in the Design Arena coding performance test conducted by human blind testing, it surpassed Claude Fabe 5 to claim the top spot in the world.

GLM-5.2, a Chinese model surpassing Claude Opus 4.7, has been officially announced. It has even outperformed Claude Fable 5 in some tests and is available for download as an open model - GIGAZINE



GLM-5.2 is available via API, and OpenRouter, which handles APIs from various AI companies, also provides a GLM-5.2 API. The image below shows the daily share of API models used on OpenRouter, with white representing Z.ai's share and blue representing DeepSeek's share. Z.ai's share has increased sharply since June 16, 2026 (local time), when GLM-5.2 was released.



OpenRouter offers several Z.ai models, but the GLM-5.2 reportedly accounts for 75% of the market share. OpenRouter commented that 'the GLM-5.2 is gaining popularity much faster than most other open models.'



Looking at the OpenRouter

model details page , the most common use of GLM-5.2 is the AI agent 'Hermes Agent,' and it is also used by agents such as 'pi,' 'Claude Code,' 'Kilo Code,' and 'Cursor.'



The official X account for the AI coding agent Cline also highly praised GLM-5.2, reporting that 'when we had GLM-5.2 and Claude Opus 4.8 fix existing bugs in Cline, GLM-5.2 was superior in terms of cost and quality,' 'GLM-5.2 consumed twice as many tokens as Claude Opus 4.8, but at half the cost,' and 'GLM-5.2 verified whether the task was compilable before completion, while Claude Opus 4.8 did not verify and left errors.'




Related Posts:

in AI, Posted by log1o_hf