Anthropic’s leading 71 percent implied probability stems from its Claude Opus and Sonnet models posting the strongest recent results on independent benchmarks such as Arena Elo ratings and real-world coding evaluations through mid-May 2026. Traders see these large language models delivering superior agentic performance, lower error rates on complex tasks, and higher user preference compared with rivals. Google’s 21 percent share reflects Gemini 3.1 Pro’s competitive multimodal and reasoning scores, while OpenAI’s 6.5 percent accounts for solid but less dominant GPT-5 family updates. With resolution set for the end of June, any new frontier model releases or updated third-party leaderboards in the coming weeks could still shift trader consensus on which company holds the top AI capabilities.