Z.ai confirms Ox Alpha is GLM-5.3-Flash, plans to release weights under MIT license
Z.ai has officially revealed that Ox Alpha, the mystery model previously launched in stealth mode on OpenRouter, is GLM-5.3-Flash, with its model weights set to be released for developers to use freely.
After sparking curiosity among developers on OpenRouter as an unidentified "stealth" model, Z.ai has officially confirmed that the model is **GLM-5.3-Flash**. The latest model in the GLM family is designed specifically for coding and agentic applications.
In terms of technical specifications, GLM-5.3-Flash is a multimodal model that supports text, images, and video. It uses a hybrid architecture combining Sparse and Linear Attention, with 320B total parameters (18B active), and supports a context window of up to 1 million tokens.
Z.ai also confirmed welcome news for the open-source community: it will release the **weights** for GLM-5.3-Flash under the MIT license, allowing developers to build on the model. This aligns with information published by zai-org on Hugging Face.
Releasing the weights of a large, high-performance multimodal model under the MIT license gives AI developers more options to deploy, innovate, and build applications at lower cost.