Seattle-based Carbon Robotics has introduced the Large Plant Model (LPM), a new AI system that powers its LaserWeeder robots, enabling real-time plant species recognition and precise weed targeting. The model allows farmers to designate which plants to eliminate without retraining the system for new weeds.
Trained on over 150 million photos and data points collected across 100 farms in 15 countries, the LPM forms the core of Carbon AI, the intelligence behind the company’s autonomous weed-killing robots. Previously, identifying a new weed required creating new data labels and retraining the machines—a process that took roughly 24 hours. With LPM, the robots can instantly identify and act on plants, even if they’ve never encountered them before.
The AI update will be delivered through a software rollout to existing machines. Farmers can now interact directly with the robot interface to label plants for protection or removal, allowing the system to refine its understanding continuously.
Founded in 2018, Carbon Robotics has raised over $185 million from investors including Nvidia NVentures, Bond, and Anthos Capital. The company plans to continue fine-tuning LPM as the robots gather additional plant data, leveraging its expansive dataset to make near-instant, accurate plant identification on the farm.
