Fireship · 2026-09-01 · notable
Fireship — 'A mysterious new model just took over the internet'
Fireship covers Ox Alpha, the anonymous model that the video says served 42 trillion tokens on OpenRouter in six days before being identified as Z.ai's GLM-5.3-Flash.
Fireship walks through Ox Alpha, the anonymous model that topped OpenRouter before it turned out to be Z.ai's GLM-5.3-Flash.
What is it?
Fireship published 'A mysterious new model just took over the internet...' on 1 September 2026. The subject is Ox Alpha, an anonymous model on OpenRouter that the video says served 42 trillion tokens in six days before it was identified as Z.ai's GLM-5.3-Flash. The weights release itself is already a separate entry in this feed; this is the channel's short breakdown of how it played out.
How does it work?
GLM-5.3-Flash is a mixture-of-experts model with 320B total parameters and 18B active, published under the MIT license. Its model card calls it the first natively multimodal model in the GLM-5 series and describes a mix of sparse and linear attention for efficiency, with Z.ai putting it at one-tenth the price of GLM-5.2. Running a model under a codename on OpenRouter is a common way to collect real developer usage before a lab attaches its name.
Why does it matter?
The anonymous launch is the interesting part: developers chose Ox Alpha on price and output quality with no brand attached, and only found out afterwards which lab had built it. Fireship compresses that into a fast explainer, which is a quicker route into the story than reading the model card and the OpenRouter leaderboard side by side.
Who is it for?
developers following open-weight model releases
Try it
Watch at youtube.com/watch?v=r-tzcMlQISk