Products

Data, environments, and evaluations for advancing real-world model capabilities.

RL Environments

Tool-rich environments where agents practice long-horizon work under realistic constraints.

Off-the-Shelf Datasets

Prebuilt and reviewed datasets ready for training across economically valuable domains.

Human & Model Trajectories

Full traces of work by experts and models, including reasoning, tool calls, and solutions.

Benchmarks

Novel benchmarks & evals built from scratch to test emergent model behaviors.

Custom

Custom Datasets & Environments

Have an idea for a dataset or environment?

We work with labs & enterprises to scale up production of new datasets in days rather than weeks.

Reach out