Anthropic’s Frontier Red Team, working with Andon Labs, published Drone-Bench, a new benchmark evaluating whether AI models can autonomously control a flying drone to locate and follow a person. The evaluation broke the task into five sub-tasks and tested 15 models, finding that Claude Fable 5 performed best overall but was held back by weaknesses in 3D reconstruction. Anthropic said newer models are steadily improving across all sub-tasks, suggesting fully autonomous drone operation could become feasible in the near future.
