Theo Shares AI Model Tier List With Opus 5 in D
Ranks 15 models from S+ to F with Fable 5 at top.
TLDR
Theo posted a tier list evaluating 15 AI models. Fable 5 received the top S+ rating. Opus 5 landed in D alongside Sonnet 5. Theo noted that Deepseek and GLM lack vision capabilities, which affected their scores. Victor Taelin replied that the list contains zero errors. Other replies questioned Grok's placement below Kimi and asked about missing models like Ox. One commenter observed that Google models were isolated in a low tier because they cost more than the rest.
Combined views
1.4M
9 Sources, first seen ago
Sources
LA 
Lisan al Gaib@scaling01the Opus 5 disrespect is insane https://twitter.com/theo/status/2091277536600969276
- likes: 1.1K
- replies: 149
- bookmarks: 112
- reposts: 9
LA Lisan al Gaib@scaling01Opus 5 is the closest thing we have to whatever Anthropic has internally it's an absolute monster if you want to optimize anything give it a target and it will hillclimb https://twitter.com/scaling01/status/2091285961585893833
- likes: 1.1K
- replies: 72
- bookmarks: 397
- reposts: 17
T- Theo - t3.gg@theoI refuse to believe these people are real https://twitter.com/vijarkohli/status/2091303019853865472
- likes: 2.5K
- replies: 234
- bookmarks: 141
- reposts: 30
T- Theo - t3.gg@theoI won’t lie, I’m surprised how much yall agreed with this one https://twitter.com/theo/status/2091277536600969276
- likes: 1K
- replies: 187
- bookmarks: 86
- reposts: 8
HH Hamel Husain@HamelHusainI was wondering why Google is in its own special toilet tier in this video 🤣 Summary: it’s expensive as hell compared to everything else https://twitter.com/theo/status/2091277536600969276
- likes: 30
- replies: 9
- bookmarks: 3
- reposts: 0
EB Ethan Breitkreutz@ebreitk@theo I like Gemini 3.7 flash for a multi modal model for classifying documents and doing OCR in my apps… it works pretty good, on my eval harness it’s usually within 1-2 points of sonnet.