Carbon Robotics launches “Large Plant Model” to expand AI-driven weeding

Carbon Robotics said it has developed what it calls the agriculture sector’s first Large Plant Model, an artificial-intelligence system designed to identify crops and weeds across a range of field conditions and geographies.

The Seattle-based company said the model was trained on a dataset of 150 million labeled plants, covering different crops, weed species, soil types, climates and growth stages. The system is designed to operate with the company’s LaserWeeder machines, which use computer vision and lasers to remove weeds without chemicals.

The model forms the core of what the company refers to as its Carbon AI platform, which runs across its product line. In addition to the LaserWeeder, the software also supports Carbon Robotics’ Autonomous Tractor Kit, a retrofit package that adds autonomous functions to existing tractors. The platform processes plant imagery and field data in real time to guide plant identification, weed removal and machine navigation, according to the company.

Carbon Robotics said its installed base of LaserWeeders supplies a continuous stream of field data that is used to update and refine the model. As more machines operate in commercial fields, additional plant images and operating scenarios are incorporated into the system.

A new feature called Plant Profiles is intended to let growers adjust how the model performs under local conditions. Using an operator app, users can add a small set of field images to modify how the system distinguishes crops from weeds in specific soils, lighting or growth stages. The company said the adjustments can be made within minutes, rather than the longer development cycles often associated with custom agricultural AI models.

Paul Mikesell, founder and chief executive of Carbon Robotics, said the objective is to enable equipment to recognize and respond to plants in unfamiliar fields without lengthy setup. A farm manager at Bland Farms, a US vegetable grower, said the feature has been used in onion production to refine performance across different planting stages.

The rollout comes as investment continues to flow into the computing infrastructure that supports artificial-intelligence applications. Semiconductor startup Positron recently raised $230 million in Series B funding at a reported $1 billion valuation, according to TechCrunch, to expand deployment of high-speed memory technology aimed at AI inference workloads. Such advances are expected to influence how quickly data-intensive models can be deployed beyond data centers and into field equipment.

For agricultural equipment makers, the shift reflects growing demand from farmers to reduce labor requirements and limit herbicide use while maintaining crop yields. Companies developing machine-vision and automation systems are increasingly positioning AI as a way to improve precision and consistency in field operations.