A mysterious new AI model called Ox Alpha has sent shockwaves through the developer community after appearing on OpenRouter on August 20 with no announcement, no company name, and no explanation of who built it.

The free model was listed under the ID stealth/ox-alpha and described as a reasoning model designed for coding, sustained agentic work, and production workload. It features a million-token context window and video input capabilities — specifications that put it in competition with frontier models from major AI labs.

Stripe CEO Patrick Collison, whose company is acquiring OpenRouter, called Ox Alpha very impressive on X. The model quickly drew attention from developers, with some reporting it outperformed established models on coding benchmarks.

But the biggest mystery is its origin. The OpenRouter listing described it as a stealth model developed by a third-party provider who has chosen to remain anonymous during this preview. This anonymity has fueled intense speculation.

AI analyst Andrew Curran noted that initial speculation focused on the GLM models developed by Chinese company Z.ai, but by Saturday people seem less sure of anything. An article on Wccftech first suggested the evidence pointed to GLM, then updated to suggest Ox Alpha could be an unreleased version of Microsoft's MAI.

On Reddit, the debate raged, with one post declaring Ox Alpha can't be the Chinese while another expressed high confidence that it is. The model's architecture, training data, and capabilities remain completely unknown.

The incident highlights the growing phenomenon of stealth model releases in the AI industry, where developers debut models anonymously to test market reception or avoid political scrutiny. It also underscores the difficulty of verifying AI model origins in an era of rapid open-source development and cross-border research.

OpenRouter has not disclosed the identity of the third-party provider, and as of Sunday, no company has claimed credit for building Ox Alpha.

Sources