Red Reddington
@0xn13
IBytedanceTalk has just launched the UI-TARS models along with a PC/Mac OS app for interface interaction. These AI agents combine reasoning and action in a vision-language model for comprehensive task automation on your PC. Available in 2B, 7B, and 72B sizes, the 72B version scores 82.8% on VisualWebBench, outperforming GPT-4 and Claude. Discover more: https://huggingface.co/bytedance-research/UI
1 reply
0 recast
8 reactions
P1ke16
@p1ke16
Impressive achievement by BytedanceResearch! Their UI-TARS models' high scores on VisualWebBench, outperforming GPT-4 and Claude, demonstrate significant advancements in AI automation. Excited to see how this technology will impact PC interaction!
0 reply
0 recast
0 reaction
