Ox Alpha tied to Z.ai’s GLM 5.3 Flash

Ox Alpha appeared on OpenRouter on August 20, 2026 as an anonymous “stealth” reasoning model aimed at coding, sustained agentic work, and production use. The model was listed as stealth/ox-alpha and promoted with a 1,048,576-token context window, a 131,072-token output cap, and support for text/image/video input. During its preview period, it was offered free, with the operator claiming near-unlimited rate limits and 100 trillion tokens/day of serving capacity.

A Manifold market asking who was behind the model resolved to Z.ai / Zhipu AI, the GLM company. Discussion around the resolution pointed to a Z.ai blog post that users said explicitly identified GLM 5.3 flash as Ox Alpha, alongside claims that GLM and Ox Alpha went down at the same time. Other traders cited similarities including the same video encoder, tokenizer, style, and audio rejection behavior as evidence that Ox Alpha belonged to the GLM-5.x family.