Z.ai Admits It Secretly Tested Its Own AI Model Under a Pseudonym
Z.ai, a Chinese artificial intelligence firm, has confirmed that the previously unidentified AI model called Ox Alpha was actually its own GLM-5.3-Flash. The company had anonymously placed the model on the OpenRouter and OpenCode platforms for testing. This disclosure ends speculation about the model's true origin.
Z.ai confirmed that the model previously circulating as “Ox Alpha” was its own GLM-5.3-Flash, submitted anonymously to OpenRouter and OpenCode for real-world evaluation. The company said the pseudonymous release was a deliberate testing strategy, allowing the model to be assessed without brand bias or prior expectations.
The admission resolves weeks of speculation among developers and AI observers who had puzzled over Ox Alpha’s strong performance and unclear lineage. By revealing the model’s true identity, Z.ai also highlights a growing trend: firms using third-party platforms to gauge competitive standing while masking their involvement.
This disclosure could affect trust in AI benchmarking, as developers may now question whether other anonymous models are corporate tests. Users and businesses relying on OpenRouter or OpenCode might face uncertainty about model provenance, potentially slowing adoption. However, it may also encourage more transparent testing practices industry-wide, as competitors and regulators could push for clearer labeling of experimental systems.