Arena Releases Week 42 AI Rankings with Domestic Models Holding Steady
Arena published its Week 42 AI model blind-test rankings for 2026, revealing that domestic Chinese models maintain strong upper-middle positions while Moonshot's kimi-k3-max broke into the top ten of the coding leaderboard. These rankings track the incremental progress of domestic AI models against dominant global competitors, highlighting specific technical domains like coding where regional models are beginning to close the gap. The update features three newly entered models—gemini-3.5-flash-medium, mimo-v2.6-pro, and flux-2-flex—with gemini-4-argon-high retaining the top spot on the overall leaderboard with an ELO score of 1525.
## BACKGROUND
Arena platform rankings rely on anonymous user blind tests and calculate positions using an Elo rating system akin to competitive gaming. This approach reflects subjective human preferences and crowd-sourced feedback rather than standardized benchmark scores.