An anonymous "stealth" reasoning model that appeared on OpenRouter on August 20, 2026 — built for coding and sustained agentic work, free during a limited preview, and widely fingerprinted to Z.ai's GLM family.
Last verified · 2026-08-24 · by Moe Ameen
Ox Alpha is a "stealth" AI model — a preview release whose developer has chosen to stay anonymous. It appeared on OpenRouter on August 20, 2026 under the ID stealth/ox-alpha, listed as a reasoning model built for coding, sustained agentic work, and production workloads, and offered free during the preview window. OpenRouter's own listing describes it as a stealth model with a developer who wants to remain unnamed during the preview. TechCrunch covered it a few days later under the honest headline "who's behind the new stealth model Ox Alpha?" — because nobody outside the lab knows for sure.
The OpenRouter listing shows an unusually large context window of 1,048,576 tokens (roughly 1M) with output in the low-hundreds-of-thousands of tokens, accepts text, image, and video input, and supports tool calling. That combination — long context, multimodal input, agentic framing — is why it drew immediate attention from developers, and Stripe CEO Patrick Collison (Stripe recently agreed to acquire OpenRouter) called it "very impressive."
On who made it, treat everything as unconfirmed. The strongest public evidence points at Z.ai's GLM family: community "fingerprinting" tools that probe a model's tokenizer and infrastructure reported close matches to GLM-5.3, and one independent tester found Ox Alpha's token counts matched GLM-5.3 across a batch of prompts with only a small constant offset. Others floated an unreleased Microsoft model. None of this is an official claim, and viral "it beats GPT-5.6 and Claude on coding" numbers came from small, unaudited user tests — read them as preliminary. The confirmed facts are narrow: it is anonymous, it is a reasoning/coding model, and it is free while the preview lasts.
The honest framing for a creator: Ox Alpha is a raw text-and-reasoning model, not a content tool. It writes code, reasons over long inputs, and drafts text; it generates no video, images, carousels, or persona avatars, and it publishes nothing. It is interesting to a creator as a very cheap drafting brain during a free window — not as a tool you would use to make and ship social content.
The useful way to think about Ox Alpha for content is opportunistic: a capable reasoning model is free right now, and free windows are exactly when you stockpile the cheap-but-tedious raw material — a month of script drafts, a hundred hook variations, outlines for every blog you have been putting off. Ox Alpha is a genuinely good drafting brain for that. What it will never do is turn any of it into something a viewer sees. It has no image model, no video, no carousel renderer, no brand layer, and no way to post. That seam — words and ideas on one side, finished on-brand posts across platforms on the other — is the entire job [Kompozy](/) does.
Concretely: draft in Ox Alpha while it is free, then bring the raw text into Kompozy as a source and let it produce the media the model can't. One drafted script becomes a captioned [Persona Short](/glossary/persona-shorts) with a face-locked avatar, a brand-exact [Carousel](/glossary/hyperframes), quote graphics pulled from the best lines, photo posts, a full blog article, and an email newsletter — 25 to 35 assets across [18 formats](/glossary/output-buckets), all held to one voice by the [Persona Brief](/glossary/persona-brief). Then [Autopilot](/glossary/autopilot) schedules and publishes the set across the eight social platforms plus blog and email, each asset clearing a per-post review gate first. A stealth model that might vanish when its preview ends is a risky thing to build on; Kompozy is model-agnostic by design, so whatever you drafted in Ox Alpha keeps flowing through the same pipeline even after the free window closes.
Ox Alpha is an anonymous "stealth" AI model that appeared on OpenRouter on August 20, 2026, listed as a reasoning model for coding and sustained agentic work and offered free during a limited preview. Its developer has chosen to remain unnamed, and community fingerprinting points strongly toward Z.ai's GLM family, though nothing is officially confirmed.
Officially, no one has said. The strongest public evidence — tokenizer and infrastructure fingerprinting plus matching token counts — points at Z.ai's GLM-5.3, and some observers speculated an unreleased Microsoft model. Treat the origin as unconfirmed until the developer claims it.
It was free on OpenRouter during its preview window, which was described as running roughly a week from the August 20, 2026 launch. Free-preview pricing on a stealth model is temporary by nature, so confirm current access on OpenRouter before depending on it.
Only the text. Ox Alpha is a reasoning and coding model — it drafts scripts, outlines, and copy and reasons over long inputs, but it generates no video, images, or carousels and publishes nothing. To turn its drafts into finished, scheduled posts you pair it with a content engine like Kompozy that makes the media and publishes across platforms.
Bring the draft into Kompozy. From that one source Kompozy generates 18 formats — persona/avatar video, carousels, quote graphics, photo posts, a blog, and a newsletter — holds a consistent face and voice across them, and schedules and publishes across the eight social platforms plus blog and email.