OpenAI’s commanding market-implied probability for second-best AI agent lab reflects its recent DevDay launch of always-on “dots” agents powered by GPT-6 Astra, which give models dedicated cloud computers and deep integration across thousands of apps including Slack and Teams. This release, combined with strong showings on agentic benchmarks such as BrowseComp and tool-use evals, has kept OpenAI in the number-two slot on the arena.ai Agent Arena leaderboard behind Anthropic’s Claude models. Traders see limited time for rivals like Meta’s Muse or Moonshot’s Kimi K3 to close the gap before the September 30 resolution. A last-minute leaderboard shift or unexpected capability demonstration from another lab remains the main realistic scenario that could alter the outcome.