Recent WebDev Arena benchmarks on arena.ai show Anthropic’s Claude Opus 5.5 and OpenAI’s GPT-6 Astra series dominating top agentic coding scores for web app generation, with Google’s Gemini 4 variants close behind on cost-adjusted performance. This fragmentation supports the closely matched market-implied odds around 45–50% for Google, OpenAI, Meta, and several others as potential third-best labs by month-end. Key differentiators include Anthropic’s edge in front-end reasoning and iterative refinement, OpenAI’s strength in full-stack tool use, and Google’s competitive pricing alongside Meta’s Muse models gaining ground through recent releases. New model drops or leaderboard updates in the final weeks could still shift lab rankings before resolution.