A Frontier-Class Model Appeared With No Name on It, and It Is Free Until Next Week
Ox Alpha turned up on OpenRouter with a million-token context and no owner. Community fingerprinting points at Zhipu's unreleased GLM, but nobody has claimed it.
An anonymous model called Ox Alpha appeared on OpenRouter on 20 August. It has a context window of 1,048,576 tokens, which is roughly a million words of working memory, accepts text, images and video, supports function calling, and costs nothing during a preview period of about a week. It is served under the model ID stealth/ox-alpha. As of 22 August, no company has publicly claimed it.
This is a stealth launch, and it is a well-established ritual. A lab wants real usage data on a model before announcing it, so it ships it namelessly through a router, watches how people push it, and then either attaches its brand to the results or quietly withdraws it. Users get free access to something good. The lab gets a benchmark run by thousands of strangers with no press cycle attached and no reputational cost if it disappoints.
Naturally, people immediately try to work out whose it is. The leading theory points at Zhipu AI’s unreleased GLM-5.3. Independent researcher Ben Davis published a fingerprinting analysis on 21 August claiming a 99 percent confidence match, based on how the model consumes video tokens and how its tokenizer segments text. Tokenizer behaviour is a useful fingerprint because it is a fixed, idiosyncratic property of how a lab chops text into pieces, and it tends to survive fine-tuning. This is community detective work, not confirmation, and it should be read as a strong hint rather than a fact. Early community benchmarks put Ox Alpha at 80 percent on the DeepSWE coding benchmark, ahead of Claude Fable 5 at 65 and GPT-5.6 Sol at 52, though early independent tests on a free preview model are exactly the kind of number that shifts once more people run them properly.
What this means for you: if you want to try a very large context window without paying for it, this is a rare window and it closes around 27 August. Paste in a whole codebase or a long document and see what it does. But there is a caveat that matters more than the benchmark: you do not know who is operating this model, where the servers are, or what happens to what you send. Free preview models are the worst possible place for client data, contracts, health information or anything covered by GDPR, precisely because the operator is unidentified. Treat it as a sandbox for public or synthetic material, enjoy the million tokens, and keep the real work on a provider whose name you can read on a contract.
Sources
Source: https://cryptobriefing.com/ox-alpha-stealth-ai-model-1m-context/
Adobe Firefly Can Now Make the Music, the Voiceover and the Sound Effects
Adobe moved Generate Music, Generate Speech and Generate Sound Effects out of beta on 20 August. All three come cleared for commercial use, which is the part that actually matters.
