Mystery solved. Ox Alpha, the anonymous model that processed 16 trillion tokens for 221,000 users in one week on OpenRouter, is GLM-5.3-Flash from Z.ai. The kicker: the whole thing ran on domestic Chinese AI accelerators. The stealth launch was a production stress test, and it passed in public.
What it actually is
The first natively multimodal model in the GLM-5 series: a 320B-parameter MoE activating just 18B per token, with text, image, and video input and a 1M-token context. Weights hit Hugging Face under MIT license on day one. It scores 57 on the Artificial Analysis Intelligence Index, and Z.ai’s coding and agentic benchmarks put it near Claude Opus 4.8 — from an 18B-active model anyone can download.
The API is the real pitch
$0.15 input / $0.50 output per million tokens — about one-tenth of standard GLM-5.3, and 50% off until September 9. At roughly $0.09 per task, it sits on the intelligence-vs-cost Pareto frontier. If you’re running coding agents or long-context video analysis, this is now the cheapest serious option with open weights as the exit hatch.
You Might Also Like
- Jupid hit 1 on Product Hunt During tax Season 507 Upvotes for an ai Accountant Built on Claude Code
- Gpt 5 5 Takes Back the Coding Crown From Claude Opus 4 7
- Zcode Zhipu z ai glm 5 2 Coding Agent hit hns 1 Spot Chinas Open Answer to Claude Code
- Claude Opus 5 Lands Within 0 5 of Fable 5 on Coding at Half the Price
- Starnus Just hit 1 on Product Hunt and Yeah its Worth the Hype
Top AI Product 
Leave a comment