Follow me on X → https://x.com/mattjohnston_ai
A secret “frontier” model just dropped with no name, no lab, and a million-token context — and the internet is already swearing Ox Alpha is GLM 5.5. I ran it through my brand-new benchmark first. Halo, XCOM, a printable TARDIS, a Diablo clone, a 5th Avenue drone flight… 91 weighted. Then I tell you why I’m still not touching it for real work.
0:00 The secret Ox Alpha model
0:20 Halo clone (94)
1:10 XCOM clone — broken but gorgeous (84)
2:54 3D-printable TARDIS phone case (92)
4:10 Sales page + animated SVGs (95)
5:44 Diablo clone — new test (95)
7:36 5th Avenue drone flight (78)
9:55 Penguin Meltdown SVG (92)
11:08 Tool calling 100, skepticism 90, final 91
12:30 What is Ox Alpha? GLM 5.5 or Gemini?
Ox Alpha is free right now. I’m not linking a mystery endpoint I won’t use for work — search it if you want to try it.
If this is GLM, it’s a lateral move vs GLM 5.3 on my bench. Next up: the new bench vs my top 20 models.
Subscribe so you don’t miss the rankings.
https://mattjohnston.io
source
