Ox Alpha: The Free Anonymous AI Model With a Million-Token Context Window

By NewSunSEO | September 16, 2026

Developers now have free access to Ox Alpha, an AI model with a context window of just over one million tokens that is drawing strong reviews for coding and long-horizon agent work. It appeared on OpenRouter last Thursday as a stealth release, offered at no cost with near unlimited usage during its first week. The one catch that matters for anyone using it on real projects: the provider behind it has not said who it is, and OpenRouter's listing states that prompts and completions are retained by that provider.

What is Ox Alpha?

Ox Alpha is an anonymous model listed on OpenRouter, published by an unidentified third-party provider. The scale on offer is large. The context window sits just above one million tokens, and the open-source agent OpenCode said the model would be free for a week with near unlimited usage, adding that its provider had capacity for 100 trillion tokens a day. It is positioned for coding, long-horizon agent tasks, and production use.

Early reaction from developers has been positive. Users testing it have called it very impressive, including praise from the leadership of the payments company Stripe. A stealth launch of this kind is a recognized way to benchmark a model against real usage before an official announcement.

Who built it?

The provider's identity remains unconfirmed. One leading theory points to Z.ai, which previously tested a model named GLM-5 anonymously under a different name. A competing analysis of the model's tokenizer suggests Microsoft's MAI family instead. By the weekend, confidence in every theory had faded, and analysts tracking the release said people seemed less certain than they had been a day earlier. The guessing continues without a firm answer.

Why the data terms matter more than the name

The more useful detail is not who made Ox Alpha but what happens to the data you send it. OpenRouter's own listing states that prompts and completions "are retained by the provider and are not used for training." In practice, whatever you submit goes to a company that has not identified itself, and that company keeps it.

For businesses in Europe, that is a hard blocker rather than a curiosity. Data protection law requires a contract with a named processor and an assessment of where data travels, and neither is possible when the counterparty is anonymous.

How do EU rules change the picture?

The timing raises the stakes. The AI Act's transparency obligations took effect on 2 August, with penalties reaching €15mn or 3% of global turnover. That framework is built on knowing which provider is responsible for what, which an anonymous release cannot satisfy.

Is the model itself a problem?

The uncertainty around the provider does not make the model low quality. Open-weight releases have closed the capability gap faster than the safety gap, and quiet benchmarking before a launch is a legitimate practice. The cost here is information rather than money. Someone is paying to serve 100 trillion tokens a day of inference, and until that provider names itself, the sensible approach for European users is to try Ox Alpha only with data that does not matter.

FAQ

What is Ox Alpha's context window?

Ox Alpha has a context window of just over one million tokens. It is positioned for coding, long-horizon agent work, and production use.

Does Ox Alpha keep the data you send it?

Yes. OpenRouter's listing states that prompts and completions are retained by the provider and are not used for training. Because the provider is unnamed, users cannot form a data processing contract with it.

Who is the provider behind Ox Alpha?

The provider has not been confirmed. One theory points to Z.ai, which previously tested GLM-5 anonymously, while an analysis of the model's tokenizer suggests Microsoft's MAI family. Confidence in every theory has faded.