Toyota Research Institute
Toyota's US research company working on robotics, large behavior models for robot manipulation, automated driving and materials, with offices in Los Altos, California and Cambridge, Massachusetts.
Toyota Motor Corporation announced on 6 November 2015 that it would set up Toyota Research Institute, Inc. as a new company for artificial intelligence research and development, starting operations in January 2016 under chief executive Gill Pratt, with a planned investment of $1 billion over five years and a target staff of about 200. Toyota said the initial focus would be AI and robotics, aimed at safer driving, mobility for more people and robots that help older people at home. TRI now lists offices in Los Altos, California, and Cambridge, Massachusetts, and says its research spans automated driving, energy and materials, human-centered AI, human interactive driving, large behavior models and robotics. In October 2024 TRI and Boston Dynamics announced a partnership to develop general-purpose humanoids by running TRI's large behavior models on the Atlas robot, and in August 2025 the two released a video of Atlas completing a long, continuous sequence of tasks that combined walking with object manipulation. In a 2026 study in Science Robotics, TRI's large behavior model team drew on about 1,700 hours of robot data, 1,800 real-world rollouts and more than 47,000 simulated rollouts and reported that pretrained models picked up new tasks with three to five times less data in demanding conditions.
- Latest funding
- —
- Total funding
- —
- Latest valuation
- —
- Founded
- 2016
- Country
- United States
- AnalysisKorea clears 20 Robros humanoids to work beside people, and the fence rule it lifts is the bottleneck Figure, Atlas and UBTech face too
- AnalysisSkild AI and Figure top the largest humanoid and embodied-AI funding rounds of 2025-2026, as money moves from robot bodies to robot brains
- IndustryWhat is a vision-language-action (VLA) model, and who is building them?
- ResearchToyota Research Institute's blind trials in Science Robotics find multitask robot models learn new tasks with a fraction of the data
- ProductNVIDIA previews GR00T N2, betting robots should predict video before they move