In a groundbreaking development, the combination of R1 as the architect and Sonnet as the editor has achieved a new state-of-the-art (SOTA) performance on the aider polyglot benchmark, reaching a 64.0% completion rate. This duo outperforms the previous best, o1, by a significant margin, achieving this feat at 14 times less cost. Interestingly, pairing o1 with Sonnet did not yield better results than using o1 alone, highlighting the unique synergy between R1 and Sonnet. This contrasts with earlier models like o1-preview and o1-mini, which benefited from being paired with various editor models. The success of R1 and Sonnet underscores the potential of combining different AI models to enhance coding tasks, offering a more efficient and cost-effective solution for developers.
Read more at aider…
