benchmax

benchmax is a python sdk for defining and training reinforcement-learning environments: tools, rewards and datasets as code.

extend BaseEnv for a simple model-and-tool loop, or use HarborEnv to train with arbitrary harnesses and sandboxes through harbor.

the castform package adds dataset generation helpers, environment validation, and launching training jobs.

python 3.12 is required.

install

uv tool install -U castform
# or
pip install -U castform

this will install both castform and benchmax

check that the cli is available:

castform --version

get started

castform setup

this signs you in and creates a starter environment.

working examples live under examples/

see the benchmax guide and the castform guide.