Z.ai Confirms Stealth Model Ox Alpha Is a New GLM Model and Will Release Weights
Chinese AI company Z.ai has confirmed that the high-performing stealth model Ox Alpha is a new iteration of its GLM series. The company announced that it will release the open weights for the model, resolving speculation about its origin. Releasing the open weights of a competitive model like Ox Alpha allows Z.ai to remain highly competitive against rivals like DeepSeek in the open-source AI ecosystem. This move provides developers with access to a powerful, cost-effective LLM capable of complex reasoning and coding tasks. Early tests show Ox Alpha utilizing a significant number of thinking tokens for complex tasks, such as successfully generating Java bindings in a single run. However, benchmark results remain mixed, with some platforms showing it outperforming Fable while others place it below GPT-5.4 Nano.
## BACKGROUND
GLM (General Language Model) is a series of open-weight large language models developed by the Chinese AI company Z.ai, also known as Zhipu. Prior to this confirmation, Ox Alpha had gained attention as a high-performing, anonymous "stealth" model available at zero cost, leading to widespread speculation about its creator.