New Models AI Minute Newsroom 2026-08-26

The nameless model that burned through 42 trillion tokens in six days has an owner: Z.ai says Ox Alpha is a GLM, and the weights are opening

Ox Alpha turned up on OpenRouter and OpenCode on 20 August with no company name attached: free to use, a 1,048,576-token context window, up to 131,072 tokens of output, and text, image and video input. It climbed to the top of OpenRouter's leaderboard, at one point running more than double DeepSeek's volume, and by the company's own count processed 42 trillion tokens in six days. On 26 August the Chinese lab Z.ai — previously known as Zhipu — confirmed it built the model, calling it a reasoning model designed for coding, sustained agentic work and production workloads, and the newest step in its GLM line after GLM-5.3. Z.ai says it is releasing the weights and keeping a week of free access open. Developers had already worked it out from forensics: one researcher ran 95 tokenizer probes against Z.ai's published GLM-5 vocabulary and matched all of them with a mean error of zero, and another deliberately sent a malformed request and got back an error message identical to the one in Zhipu's own API documentation.

Why it mattersWe ran this as an unverified rumour on 25 August, on the strength of a tokenizer fingerprint and a stack trace, and wrote that whoever built it would put their name on it within a couple of weeks. It took one day. The confirmation matters more than the gossip did: it means the fingerprinting techniques developers now use on anonymous endpoints actually work, and that a lab can no longer run a stealth preview and count on staying anonymous. It also means a model that spent a week at the top of a major routing marketplace, free, was taking everybody's prompts through a provider nobody could name — a provider that now, at least, has an address.

#Coding #AI Agents

✓ Verified · 4 sources

Claude's memory now follows you from the chat window into the work app — and it is built to forget your politics unless you say otherwise
2026-08-26 The Qwen4 preview is open: six billion parameters active per token, and on Alibaba's own table it fixes more real bugs than Opus 4.6
2026-08-26 An agent that keeps thinking after you stop typing — and the whole harness is under 10,000 lines of Bash
2026-08-26 Show the robot one video of a ten-minute job, and it does the ten-minute job — no retraining, nothing typed in
2026-08-26 Alibaba has put a countdown clock on the first piece of Qwen4 — the weights open tonight at 23:00 in Beijing
2026-08-26

The nameless model that burned through 42 trillion tokens in six days has an owner: Z.ai says Ox Alpha is a GLM, and the weights are opening

Related stories