Model card

Made by
Ai2
Size
5.5B parameters
Published
4 May 2026
More about the model
Publication
arXiv 2605.02881
Weights licence
Apache 2.0, for research and education under Ai2's Responsible Use Guidelines source
Actions per chunk
15, at 15 Hz source
Smallest GPU
Under 16 GB of GPU memory, in bfloat16 source
Architecture
Molmo2-ER vision-language model and a flow-matching action expert that reads its key-value cache at every layer source
Training data
Fine-tuned on 74,604 filtered DROID episodes (17.8M frames) source

Run with Positronic

Start the server source

docker run --gpus all -p 8000:8000 positro/molmoact2-droid

Drive a DROID robot from a positronic checkout guide

uv run --extra hardware positronic-inference real \
  --policy=.remote \
  --policy.address.host=localhost --policy.address.port=8000 \
  --next_task.instruction="Put the marker on the plate"

Progress through an episode

Each band shows how many episodes of MolmoAct2 had succeeded or had no success at each second of simulated time. An episode that has ended stays in its band. The dashed line shows the mean Reliability, the share of the episodes that had succeeded.

Episodes

Each episode of MolmoAct2 on MolmoSpaces-320, the newest first.