This Unknown AI Model is Shockingly Good

TL;DR

RC's open-source Trinity Large Thinking model benchmarks competitively against Claude Opus and Kimi K2.5 despite the company being virtually unknown.

Key Points

• 1.RC's Trinity Large Thinking is a new open-source model under Apache 2.0. This is the presenter's first time hearing of the American company RC, and the model is benchmarked against Claude Opus 4.6, Kimi K2.5, GLM5, and Minimax M2.7, performing on par with most.
• 2.The model demonstrates coding and agentic capabilities in demos. Promotional footage shows Trinity Large Thinking building a snake game, performing multi-step agentic work, and generating code at high speed, though the presenter questions whether the speed shown is real-time.
• 3.The presenter admits struggling to find meaningful benchmarks for everyday users. Rather than raw performance metrics, he wants a benchmark focused on real-world business and personal use cases, and is openly soliciting collaborators via X or LinkedIn to build one.