Recent leaderboard snapshots on Arena.ai show Anthropic and OpenAI models holding the top two lab positions in Code Arena WebDev, leaving third place contested amid rapid releases and shifting Elo scores on agentic web development tasks. Google’s Gemini 4 Argon, Alibaba’s Qwen3.8 series, Moonshot’s Kimi K3, and Meta’s Muse models sit in a tight cluster behind them, with performance edges driven by context length, tool use, and cost efficiency rather than raw scale. Chinese labs have narrowed gaps through frequent updates and competitive pricing, while trader-implied odds remain fragmented below 25% for any single contender. New model drops or benchmark volatility through year-end could reorder the third slot before resolution.