Understand this story
On November 6, 2026, Ai2 released ToyBlocks, an open-source robot learning simulation framework. It is the first controlled world in its FoundationWorlds series, built around a robotic arm and tabletop blocks. The framework uses reinforcement learning and motion planning to automatically generate tasks and demonstration data, can execute in parallel at up to 24 times real-time speed on a single GPU, and provides evaluation metrics such as execution speed, applied force, and task progress. The first data release contains 85,336 long-horizon demonstrations, nearly 1,600 hours of robot execution data, and 144 million frames, and it openly releases the simulation assets, task and demonstration generators, automatically generated demonstrations, two trained models, training code, evaluation code, and a validation set.
AI-generated from public reports
Latest development · Report
Ai2 Releases ToyBlocks Robot Learning Simulation FrameworkDevelopments
1 confirmed developments
- Source:Ai2 Blog
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Nov 6
Ai2 BlogFirst-party
Ai2 Releases ToyBlocks Robot Learning Simulation Framework
Ai2 has released ToyBlocks, an open-source simulation framework centered on robotic arms and tabletop blocks, and the first controlled world in its FoundationWorlds series. It uses reinforcement learning and motion planning to automatically generate tasks and demonstration data, can run in parallel on a single GPU at up to 24 times real-time speed, and provides evaluation metrics such as execution speed, applied force, and task progress. The first data release contains 85,336 long-horizon demonstrations, nearly 1,600 hours of robot execution data, and 144 million frames, with simulation assets, generators, and training and evaluation code all released openly.
Ai2 BlogRelated discussion
Ai2 Blog joined this discussion
This post contributes to heat; it is not a formal report or confirmed fact.
Hugging FaceFirst-party
ToyBlocks: A Controlled World for Robot Learning and Evaluation
Allen AI has launched ToyBlocks, a controlled robot learning and evaluation environment built on FoundationWorld, featuring four characteristics: realistic physics and appearance, large-scale automatic generation of tasks and demonstrations, a combination of reasoning and dexterous manipulation, and broad coverage of domain tasks. It uses reinforcement learning and motion planning to generate skills in Isaac Lab, and its task and motion planner (TAMP) can execute hundreds of plans in parallel on a single GPU, reaching up to 24 times real-time speed. Users can also connect frontier models such as Claude through a web interface and instruct the robot with natural language commands.