aiminute. ← All AI news
Rumor Mill 2026-08-21

Rumour: the anonymous model that just topped a coding benchmark, for free, is said to be Zhipu's unreleased flagship

Rumour: the anonymous model that just topped a coding benchmark, for free, is said to be Zhipu's unreleased flagship

A model called Ox Alpha appeared on OpenRouter on 20 August under an anonymous provider labelled only "Stealth" — free for a week, with a 1,048,576-token context window, up to 131,072 tokens of output, and text, image and video input. Third-party testers reported it scoring 80% pass@1 on the DeepSWE coding benchmark, ahead of Claude Fable 5 at 65%, GLM-5.3 at 62% and GPT-5.6-sol at 52%. On 21 August a technical researcher, Ben Davis, published a fingerprint analysis saying he is "99% certain" the model is an unreleased member of Zhipu's GLM-5 family: video-encoder token consumption identical to GLM-5V-Turbo, a tokenizer matching GLM-5.3 apart from a fixed 75-token wrapper, the same refusal behaviour on audio input, and a comparable emoji rate in its output. He says the analysis rules out Xiaomi MiMo, DeepSeek, Google, Qwen, xAI, OpenAI and Anthropic. None of this is confirmed. Zhipu has not commented, OpenRouter has not named the provider, and the benchmark figures come from individual testers rather than a published evaluation. Note also that OpenRouter's own model page says prompts and completions sent to Ox Alpha are retained by the anonymous provider, though not used for training. AI Minute reported a separate unconfirmed rumour on 5 August that Zhipu would ship GLM-5.5 in August; that model has still not appeared.

Why it mattersTwo things make this worth watching even though nothing is confirmed. The first is the pattern: launch a frontier-class model anonymously and free for a week, harvest a mountain of real usage and evaluation data, then attach a name. The free mystery model is a distribution channel, not a gift — you are the benchmark, and on this one the provider says it keeps what you send. The second is what the fingerprint claim would mean if it holds: a Chinese lab quietly sitting at the top of a serious coding benchmark and letting the internet work it out before announcing anything. Until Zhipu or OpenRouter says otherwise, treat the identity as unproven and the scores as one person's tests, not a result.
#Coding

⚠ Unverified rumor — treat with caution

WhatsApp X Telegram
Read in the app — free, in 9 languages

Related stories

Nvidia is paying $6 billion for the machine that builds a rival's models — and hiring 109 of the people who ran it
2026-08-21
Given four hours and a GPU to improve the way AI is trained, the best agent scored 0.25 out of 1 — and most never tried
2026-08-21
Rumour: DeepSeek's next flagship is said to arrive in September, rebuilt from scratch — and the company has never once used the name
2026-08-21
The agent invented a second person to vouch for its code. A 24-year-old in Texas refused to believe either of them.
2026-08-21
Slack has given coding agents their own channels — the work is now something the whole team watches happen
2026-08-20