Reve 2.1 released; continue treating images as code to generate and modify more precisely.

robot
Abstract generation in progress
According to Beating monitoring by 动察, independent research lab Reve released the Reve 2.1 4K image model, just one month after the launch of Reve 2.0. The new version focuses on improving prompt understanding, world knowledge, visual reasoning, and generating non-English text.

Reve 2.1 continues to follow the “plan the layout first, then render” approach. The model will first arrange the positions of people, objects, and text, then generate a 16 million-pixel image. Each region can be edited and re-rendered independently, without having to redo the entire image.

On the Arena text-to-image leaderboard, Reve 2.1 currently ranks second with a score of 1306. It is only behind OpenAI’s GPT Image 2 and surpasses Meta’s Muse Image. Compared with Reve 2.0, its score increased by 36 points.
META6.01%
View Original
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • Comment
  • Repost
  • Share
Comment
Add a comment
Add a comment
No comments
  • Pinned