Z.ai has confirmed that Ox Alpha, a high-profile model on OpenRouter, is a new iteration of its GLM series, according to a Techmeme pickup of Bloomberg reporting by Luz Ding. The same report says Z.ai plans to release the model’s weights tonight. The confirmation follows a short period in which Ox Alpha circulated as a mystery model. A Dejan.ai post surfaced on Hacker News described Ox Alpha as an undisclosed large language model that had appeared on OpenRouter and was gaining attention on the leaderboard. The post said some users initially speculated it could be related to Gemini, but its author argued that technical probes pointed instead toward GLM and Z.ai. The strongest new fact is Z.ai’s reported confirmation. The Dejan.ai analysis cited prompt-injection behavior, gzip-NCD compression analysis, parametric memory tests, and factual probes as evidence for the GLM link. Bloomberg’s report, as summarized by Techmeme, moves the identity question from inference to company attribution. The weights-release claim is still single-source within this cluster. Techmeme’s summary says Z.ai will release the weights tonight, but the provided materials do not specify license terms, file location, model size, or whether the release will include training details, safety documentation, or deployment constraints. OpenRouter is central to the story because it was where Ox Alpha gained visibility. Bloomberg, via Techmeme, says the model topped OpenRouter’s leaderboard. The earlier Dejan.ai post said it was climbing the leaderboard, which is directionally consistent but less specific than Bloomberg’s account. For builders, the immediate question is what “release the weights” means in practice. If the release happens as reported, independent evaluation can move beyond leaderboard placement and prompt-based identity tests toward direct inspection, local deployment experiments, and reproducibility checks. The provided sources do not yet establish performance numbers, benchmark details, or comparisons against specific rival models. Who benefits: Z.ai benefits from converting a mystery-model attention cycle into visibility for its GLM series. Developers and evaluators benefit if the weights are actually released and usable for independent testing. Who's exposed: Platforms and users relying on anonymous model listings are exposed to uncertainty around provenance until providers disclose identity and terms. Any rival model speculation around Ox Alpha is weakened by Z.ai’s reported confirmation.