The listing describes it as "a reasoning model designed for coding, sustained agentic work, and production workloads," built by a third-party provider that has chosen to stay anonymous during the preview.
Myriad: When will OpenAI release GPT-6? Click to make your prediction.What's more, the mystery AI model already tops Anthropic's Claude Fable 5 and OpenAI's GPT-5.6 Sol on impressive coding benchmarks.
DeepSWE—a benchmark that scores how often a coding agent resolves real GitHub issues correctly on its first attempt, measured as a percentage of tasks passed across 113 tasks pulled from 91 repositories—shows that Ox Alpha is competitive against the best models.
Developer Ben Davis ran an initial 10-task sample and got Ox Alpha to 80% on the first pass, ahead of Claude Fable 5's 65% and GPT-5.6 Sol's 52%. That number went viral within a day, and versions of it are still circulating as if Ox Alpha has a clean edge over both models.
I ran this thing through 10 tasks on DeepSWE (so there could be a ton of variance in it's real score, this is a subset), but uh...
gpt-5.6-sol: 52%fable: 65%whatever the hell this is: 80% (was a near miss on the "x"s so actually over 80%)
As of August 22, Ox Alpha still had no entry on Artificial Analysis or LMSys Arena.
The model is free to use, reads up to roughly 1 million tokens in a single context window, and takes text, image, and video as input while returning text. It also supports tool and function calling, though its JSON output isn't schema-enforced—a real gap for developers building agents that expect structured responses.
Such numbers make it easy to conclude that the model is so good they expect a lot of usage and the provider has enough infrastructure to support the load.
Ox Alpha (stealth model) is free for the next week
- 1M Context- Multi-modal- Zero Data Retention
Generous rate limits, near unlimited usage
We have capacity for 100T tokens per day, lets see what you can do
Why everyone's paying attention$ ori --model stealth/ox-alpha
Xiaomi's MiMo v2.5 accepts audio input, a headline feature Ox Alpha flatly rejects. DeepSeek has never shipped video capability and releases open weights instead of running stealth previews. Google and Qwen both use different tokenizers and video encoders entirely, so it’s hard to establish a solid link to those models.
What's left points to Zhipu AI's family of models. Independent fingerprinting matched Ox Alpha's tokenizer to GLM-5.3 on every normalized test and its video-token spend to GLM-5V-Turbo exactly across four separate clips, and it shares GLM's distinctive errors and audio-rejection behavior.
Zhipu AI has said nothing publicly about Ox Alpha either way.
What's a stealth model, anywayA stealth model is a frontier-class AI system released without attribution through a routing platform like OpenRouter, letting a lab collect real-world usage data before committing to a public launch.
Ox Alpha is live now on OpenRouter under the model ID stealth/ox-alpha, and through OpenCode, Cline, and Nous Portal, all free to query. OpenCode's launch note pegged the free window at about a week from the August 20 debut, putting the likely cutoff around August 27—after which, if the pattern from the last four anonymous models holds, someone may finally put a name to it.
