Ant Bailing releases trillion-level flagship thinking model Ring-2.6-1T

robot
Abstract generation in progress

Deep Tide TechFlow News, May 9 — Ant Bailing today released the trillion-level flagship thinking model Ring-2.6-1T, designed specifically for complex real-world task scenarios. The model innovatively introduces a tunable Reasoning Effort mechanism, supporting two levels of reasoning strength: high and xhigh. In real task execution evaluations, PinchBench scored 87.6, higher than GPT-5.4x High, Gemini-3.1-Pro high, and Claude-Opus-4.7x high; on highly difficult reasoning tasks, ARC-agi-V2 scored 77.78, placing it at the same level as Gemini-3.1-Pro high and Claude-Opus-4.7x high. (Jin10)

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
  • Pin