OpenAI's GPT-5 with high reasoning effort leads the Aider Polyglot coding benchmark with a score of 88.0%, establishing a clear advantage over competitors like o3-pro (84.9%) and Gemini 2.5 Pro (83.1%). This real-world benchmark evaluates model performance by editing code across multiple programming languages within existing codebases, rather than generating isolated functions.
- GPT-5 with high reasoning effort achieves 88.0%, while the medium setting scores 81.3%, highlighting a significant 6.7 percentage point gain from increased compute.
- OpenAI dominates the top six positions on the leaderboard, demonstrating depth across its model lineup.
- Claude Sonnet 4 trails significantly at 61.0%, potentially due to difficulties with Aider's specific diff-based edit workflow rather than general coding capability.
The results indicate that GPT-5 is the superior choice for teams using Aider as their primary coding assistant, while the gap between reasoning settings suggests compute scaling remains a critical lever for performance.