Every "best AI model" ranking you've scrolled this year is measuring the wrong thing, and Contra Labs just proved it with 600 blind landing pages.
Opus 5 finished last on head-to-head. 31.3% win rate. GPT-5.6 Sol won 65.3%. Case closed, right? Not quite. When the same pages
It's incredible. Took a few back-and-forth, and dealing with what I'm sure is insanely high API load causing stream errors.
Nonetheless, this may be my new favorite? Beats Opus 5...?
An interactive web page was made for Xixia Mausoleum. This dynasty
An interactive web page was made for Xixia Mausoleum. This dynasty, which had a strong history of two hundred years including the Song, Liao and Jin Dynasties and created its own writing, had nine wooden imperial mausoleums burned down and weathered into the mounds at the foot of Helan Mountain today.
Qwen 3.8 27B can now run locally on RTX 4060 8GB?!
This is what made me stop scrolling. The 27B model which usually sounds "too big for a small GPU" can actually be forced to run on an RTX 4060 with only 8GB VRAM using a new quantization from Unsloth: IQ4_XS.
You have reached the end of the archive
All of Claude Opus 5