According to Code Arena’s WebDev leaderboard updated on August 6, 2026, the performance gap between proprietary and open-weight AI models has narrowed from roughly 150 Elo points to just 11 points. Anthropic’s Claude Opus 5-max leads with a score of 1686, followed closely by Moonshot’s open-weight Kimi K3-max at 1675. Chinese AI labs dominate the top rankings, with the Kimi K3 series, GLM-5 family, and various Qwen variants all appearing in the top 10. This convergence carries significant economic implications as organizations can now self-host near-equivalent alternatives without licensing fees or API usage costs. The leaderboard is based on 531,553 pairwise human votes across 111 models.
Source: Read the original article
