OpenRouter · 2026-08-20 · major
Ox Alpha — an anonymous 1M-context coding model, free on OpenRouter
Ox Alpha is a reasoning model that appeared on OpenRouter under the provider name Stealth, with no lab claiming it. It has a 1,048,576-token context window, takes text, images and video, and costs nothing during the test window.
A free reasoning model with a 1M-token context window, live on OpenRouter with no maker attached.
Quick facts
| Maker | Unknown — listed under the "Stealth" provider |
|---|---|
| Model ID | stealth/ox-alpha |
| Context window | 1,048,576 tokens |
| Max output | 131,072 tokens |
| Input types | Text, image and video |
| Price | Free (zero per prompt and completion token) |
| Data policy | Prompts and completions kept by the provider, not used for training |
What is it?
Ox Alpha appeared on OpenRouter on August 20, 2026 under the provider name "Stealth" — no company attached and no lab claiming it. OpenRouter describes it as a reasoning model designed for coding, sustained agentic work, and production workloads, suited to long-horizon software engineering and workflows that combine text with visual context. It accepts text, images and video, and returns text.
How does it work?
Requests go to the model ID stealth/ox-alpha through OpenRouter's normal API. OpenRouter states it routes requests to the model and is not its developer, owner, or provider. Function calling works through the tools and tool_choice parameters, and JSON output through response_format without JSON-schema enforcement, so existing agent code runs unchanged.
Why does it matter?
A 1,048,576-token context window and a 131,072-token maximum output, at zero cost, make long-running coding jobs cheap to test — the kind of workload that normally runs up a bill fast. The trade-off is trust: the provider keeps prompts and completions, and nobody outside knows which lab is on the other end. OfficeChai reports that a developer's tokenizer comparison lined Ox Alpha's tokens up almost exactly with GLM-5.3, but writes that nobody involved has confirmed anything.
Who is it for?
developers testing long-context coding agents
Frequently asked questions
- Who made Ox Alpha?
- Nobody has said. Ox Alpha is listed on OpenRouter under an anonymous third-party provider called Stealth, and OpenRouter states plainly that it routes requests to the model and is not its developer, owner, or provider. OfficeChai reports that a tokenizer comparison lines Ox Alpha up with GLM-5.3 from Z.ai, and names Xiaomi's MiMo team as the other suspect, but no lab has confirmed it.
- How much does Ox Alpha cost?
- Nothing right now. OpenRouter lists Ox Alpha at zero cost for both prompt and completion tokens. OfficeChai reports the model is free for a week through OpenRouter and OpenCode. Anonymous models on OpenRouter are normally free while they are being tested, so treat the price as temporary rather than a standing rate.
- Does Ox Alpha train on the prompts I send it?
- No, according to OpenRouter's data policy for the model. The Ox Alpha page states that prompts and completions are retained by the provider and are not used for training, with all other use governed by the Stealth Model Terms. Retention still means an unnamed party holds your text, so keep sensitive code out of it.
- Where can I use Ox Alpha?
- Ox Alpha is reachable through the OpenRouter API using the model ID stealth/ox-alpha, which means any tool that already speaks to OpenRouter can point at it. OfficeChai also reports it is available directly inside OpenCode for terminal coding. The listing supports tool calling and JSON responses, so agent harnesses need no special handling.
Try it
Model ID: stealth/ox-alpha on OpenRouter