Ox Alpha: What We Know So Far

OpenRouter put a nameless model on the board on August 20, 2026, and by the next morning X had decided the interesting part was who built it. The listing calls it Ox Alpha (stealth/ox-alpha); the lab still hasn't walked out to claim it.

It's free, it takes a million tokens of context, and the activity graph already shows about 173 billion prompt tokens. That combination does not stay quiet.

Key Takeaways:

• Ox Alpha landed on OpenRouter on Aug 20 as a free stealth model with a 1M context window. As of Aug 21, no lab has put its name on it.
• OpenRouter only routes the requests. Prompts sit with an anonymous provider that says it will not train on them, which is not the same as nobody being able to read them.
• The X thread is mostly identity: next GLM with vision, or a stolen Claude checkpoint. Neither claim is on the card.
• Real traffic showed up anyway. Claude Code had already burned 36.2B tokens on it when we looked; Hermes Agent was next.

What the listing actually says

The OpenRouter card is the only first-party page. It pitches Ox Alpha for coding and long agent jobs, including ones that mix text with pictures or video. We didn't run it; this is a read of the card and the thread, not a review.

OpenRouter's own post used louder language ("frontier," "real-world production use") and a card that says CONTEXT 1.05M, released Aug 20, 2026. Around 5:02 AM on August 21 that post was sitting near 522,000 views.

You can skip the rest of the spec sheet. Context is 1M, max output is 131K, inputs are text, images, and video, and the price is $0 per million tokens in and out. Latency on the page will have moved by the time you click through.

A follow-up from the same account is the sentence people will paste into Slack: "It is free," and "This time, the provider does not train on your prompts or completions."

The naming fight

@thdxr posted the card and asked people to name the model. @theo asked too. @Suhail sounded surprised. Digg scooped the thread.

@teortaxesTex did the longest public guess, and it is still a guess. After more testing they said they can believe it's the next GLM with vision, that Zhipu sat on GLM-5.3 for a few weeks, and that it's faster than they expected in places Chinese models usually lag (they named Kimi K3). They also said they could half-believe a stolen Claude checkpoint. Speed and cache behavior look like GLM to them; DeepSeek, they wrote, it clearly isn't. Their working names are GLM 5.3-Vision or 5.5, with a rough 61 on Artificial Analysis. We haven't run those evals.

If a thread tells you "Ox Alpha is GLM," that person is theorizing.

One joke, because it is in the screenshot above and it will get quoted as analysis: @PeerReview wrote "The model is three Gemini flashes wearing a trenchcoat."

Who is already using it

The Apps tab is the part of the listing that isn't marketing copy.

Claude Code on a free mystery model is the line that travels. Hermes Agent is second. ZCode is Z.ai's own coding app, which is why the GLM guess keeps getting air even with no name on the card. Oh-My-Pi and DeepSeek Harness are in there as workers.

Logs

OpenRouter is not the lab. The banner says a third-party provider runs Ox Alpha and has chosen to stay anonymous during the preview.

They keep prompts and completions, and they say those logs are not used for training. "This time" in the thread note is doing work, because earlier stealth drops were more candid about using logs to improve the model. Kept still means kept. If you would not send a private repo to an unnamed lab, a free 1M window does not fix that.

This is not the first Alpha

OpenRouter has done nameless Alpha drops before. Pony Alpha in February 2026 later turned out to be Zhipu / Z.ai GLM-5; OpenRouter's own Pony page now says so.

Owl Alpha was later reported as Meituan's LongCat-2.0-Preview. Meituan's LongCat account said that in June 2026, and Decrypt repeated it. That was the lab naming itself, not OpenRouter unmasking anyone.

Pony to GLM-5 we can pin. Owl to LongCat we can report. The last few named Alphas were Chinese labs. None of that makes Ox Alpha the next GLM. As of August 21, 2026, the owner is still unnamed.

The 1M window still needs a machine that stays on

A million tokens of context is not a job that survives you closing the laptop. Paste-into-playground dies when the tab dies, which is the same hole we keep hitting when we look at long-horizon agent evals.

Keep the tools you already pay for. Put the long job on a machine that stays up.

FAQ

What is Ox Alpha?

A stealth model OpenRouter listed on August 20, 2026 as stealth/ox-alpha. The card says it's for coding and long agent jobs, with a 1M-token context window. The developer is anonymous during this preview.

Is Ox Alpha free?

On the listing as of August 21, 2026, yes: $0 per million tokens in and out. OpenRouter's own thread note also says "It is free." Free does not mean the traffic isn't being logged.

Who made Ox Alpha?

Unclaimed as of August 21, 2026. OpenRouter routes requests and is not the developer. Community guesses (next GLM with vision, a Claude checkpoint) are guesses.

Does the Ox Alpha provider train on prompts?

The listing and OpenRouter's follow-up both say no. Prompts and completions are still kept by the provider. Other use is under the Stealth Model Terms.

What is using Ox Alpha already?

On the listing's Apps tab as of August 21: Claude Code (36.2B tokens), Hermes Agent (28.8B), Oh-My-Pi (26.1B), DeepSeek Harness multimodal-bridge (23.7B), and ZCode (18.4B). About 173B prompt tokens show on the activity card.