Google’s late-September launch of Gemini 4 Argon (High) has elevated the lab’s math-specific Arena scores to the overall leaderboard lead, tightening the race for third place behind Anthropic’s cluster of Claude variants. Meta’s Muse Spark 1.3-max, Alibaba’s Qwen3.8-max, and Z.ai’s GLM-5.3 series sit within a narrow Elo band, while Thinky’s open-weight inkling model and DeepSeek’s v4 releases keep Chinese and smaller labs in contention through cost-efficient gains. Rapid iteration—evident in weekly Pareto-frontier shifts—means any frontier release before year-end could reorder the pack, leaving trader consensus split across a dozen labs with no dominant third-place frontrunner.