A frontier model appeared on OpenRouter on Thursday night with a 1-million-token context window, text, image, and video input, and a price of exactly zero. Nobody has said who built it, and that silence is the most interesting part of the story.
The model is called Ox Alpha, listed under the ID stealth/ox-alpha. OpenRouter is explicit that it routes requests to the model but is not its developer, owner, or provider. The operator behind it is anonymous, and the listing positions it as a reasoning model built for coding, sustained agentic work, and production workloads. It has a 1,048,576-token context window, a 131,072-token output limit, and tool and function calling with structured JSON output. That is the full agentic checklist, and the video input is the standout: none of the earlier anonymous releases advertised it.
It is free for one week, with the operator claiming generous rate limits, near-unlimited usage, and a serving capacity of 100 trillion tokens a day. Taken literally, that would put the operator among the largest inference providers anywhere, which reads less like a spec and more like a stress-test challenge. There is no known price after the free week, and there are no benchmark scores of any kind. The word “frontier” is the operator’s own claim.
What exists instead is traffic. Within the first hours, the platform’s activity data showed real production usage from Hermes Agent, the agentic coding project from Nous Research, and the Zed editor. Those are exactly the sustained-agentic-work and coding use cases the model is positioned for. It is not a score, but it is a signal: developers with real workloads decided a free, anonymous, frontier-class gamble was worth wiring up.
Ox Alpha is the fifth anonymous release in six months, and the previous four all resolved the same way: an anonymous debut, a burst of free traffic, then a company stepping forward. Pony Alpha was confirmed by Zhipu AI as GLM-5. Hunter Alpha was revealed as Xiaomi’s MiMo-V2-Pro. Elephant Alpha was claimed by Ant Group as Lingxi Ling-2.6-flash. Owl Alpha was confirmed by Meituan as LongCat-2.0 on June 30. The pattern is consistent enough that the standard reading is the lack of a brand is the marketing: anonymity removes brand bias from evaluations, and free usage crowdsources real-world evaluation data at frontier scale while everyone argues about the owner.
The one difference worth noting is the data-retention posture. The free-week announcement touts zero data retention, but the model’s listing tells a more qualified story: prompts and completions are retained by the provider but not used for training, under the platform’s stealth-model terms. That distinction matters if you are about to paste proprietary code into a 1-million-token window owned by an anonymous operator. Treat “zero data retention” as marketing until the operator publishes terms that say so in so many words.
As of this writing, no company has claimed Ox Alpha, there are no independent benchmarks, the real price is unknown, and the architecture and parameter count are undisclosed. The video input narrows the field toward labs with serious multimodal and video work, but a prior is not evidence. Until a company steps forward or the model leaks an architecture tell, every guess is a guess.
The honest verdict is that Ox Alpha is an unusually cheap experiment in both directions. OpenRouter tests whether an anonymous frontier-class model at zero dollars can win real production traffic in a week, and developers get to test whether the model deserves any. By the time the free week ends, the odds are good we will know who made it, what it costs, and whether it earned a place in a production stack. Those three answers are what make the audition worth having.
Sources: OpenRouter, OpenRouter Ox Alpha, OrcaRouter analysis, Artificial Watch