Google and OpenAI trade within a point of each other as the implied frontrunners for second-best AI agent lab, reflecting their comparable strength in agentic benchmarks such as tool use, multi-step reasoning, and computer-use evaluations. Google DeepMind’s recent Gemini 4 Hydrogen release emphasizes faster agent workflows for research and coding, while OpenAI’s GPT-6 series maintains competitive scores on agent leaderboards. Meta and Anthropic sit further back, with Meta gaining on open-weight releases and Anthropic leading many pure agent metrics but appearing priced as the probable first-place option. Chinese labs and others remain low-probability due to narrower demonstrated agent capabilities at scale. Resolution hinges on December benchmark updates and any final model drops before year-end.