Open Source RoboticsRobots, software & learning

MolmoBot

Software

Robot manipulation policies and evaluation code paired with simulated training data and the MolmoSpaces generation engine. [1] [3]

Installation & first result →Official websiteSource codeGetting started

For
Robot learning researchers comparing simulation-trained manipulation policies. [1]
Hardware
Documented DROID route [1] [3]
Source rights
Named licenses are documented by part. Review licenses
Unknowns
operating system not fully established. Review gaps

Source researched · Not tested by us
Sources checked · Research details

Licenses & related hardware

Requirements · Research details and dates

On this page

Overview

MolmoBot is a released route from procedurally generated manipulation episodes to trained robot policies. The authors provide policy code, a linked MolmoSpaces generation engine, a dataset and model cards. That makes it a useful research starting point, but each artifact has its own setup and rights. The paper's zero-shot transfer results are author-reported; this catalog has not reproduced them. [1] [2] [3] [6] [7]

Is this for you?

Workflows
  • Select policy and matching checkpoint
  • Run simulation demo or configure DROID policy server/client

[1]

Known limitations
  • Paper performance is author-reported and not reproduced here.
  • DROID evaluation needs matching hardware, cameras, Polymetis servers and checkpoint.
  • Dataset scenes carry object-specific reuse considerations.

[1] [3] [7]

What you need

Languages / runtimes
  • Python

[1]

Dependencies
  • MolmoSpaces for data generation
  • DROID/Polymetis for documented physical evaluation

[1]

Host computer

GPU resources depend on policy variant and task; no independently measured minimum.

[1]

Compute for the workload
  • Training and evaluation compute depend on variant and simulator; no verified minimum in this review.

[1]

Installation & first result

Open the upstream getting-started guide [1] · Project documentation [1]

  • Follow the policy README for the simulation demo.
  • For DROID, configure the matching camera and Polymetis services before starting policy server and evaluation client.

[1] [3]

Installation

Variant-specific README instructions; no installation performed.

[1]

Hardware, models & data

Supported hardware
  • Documented DROID route; paper also reports Franka and Rainbow Robotics platforms.

[1] [3]

Inputs and outputs

Camera observations and robot states to manipulation policy actions; exact normalization depends on selected policy/checkpoint.

[3]

File formats
  • Dataset episode packages
  • Model checkpoints

[3] [7]

Algorithms and models
  • MolmoBot
  • MolmoBot-Pi0
  • MolmoBot-SPOC

[1]

Evaluation

Simulation demo and DROID evaluation routes documented upstream; neither reproduced here.

[1]

Project notes

Choose a starting workflow

For an initial software check, enter the repository's MolmoBot/ package directory, run the documented uv sync --extra eval installation step, and follow its simulation demo with the allenai/MolmoBot-DROID checkpoint. The separate DROID route needs Polymetis robot and gripper servers, one wrist and one external camera, a policy server started with --action-type joint_pos, and a robot evaluation client. Those commands assume the matching setup; source availability does not make another arm compatible. None were run here. [1] [3] [6]

The data card describes eight task configurations across Franka and Rainbow Robotics platforms and offers bulk-download and streaming examples. It also provides an episode index for scenes whose object licenses are commercial-use friendly; the dataset-level ODC-BY label does not erase asset-specific restrictions. Model-card access was checked, but checkpoint binaries and dataset shards were not downloaded or run. [6] [7]

This entry covers the policy and its documented starting workflow. MolmoSpaces and MolmoBot-Data remain linked, separately scoped artifacts; their contents and dependencies need their own full review before standalone entries. No simulation, training, inference, DROID execution or physical manipulation was tested here. [1] [3] [7]

What we checked

Source research only. No commercial relationship or hands-on testing. No upstream media copied to the public site.

Not yet established

  • Requirements: operating system

Updates

Upstream last updated
Not established
Latest release
Not established
Profile updated
Nov 7, 2026
Sources last checked
Nov 7, 2026
Next review
Dec 7, 2026
Reviewed revision
Commit snapshot 33c0ca77bf6062a23d60ffd4a6859334c4a46d30; no matching GitHub release found

Source files and licenses

Inspectable robot-specific policy implementation with simulation and DROID workflows; data-generation engine, weights and dataset retain separate scope and conditions.

LayerAvailability / license
softwareavailable · Apache-2.0
Artifact and declaration inspected at the cited scope. External assets/dependencies and actual operation need separate verification. [3] [2]
model weightsavailable · Apache-2.0 card declaration
Artifact and declaration inspected at the cited scope. External assets/dependencies and actual operation need separate verification. [6]
datasetsavailable · ODC-BY card declaration; object-level restrictions require filtering
Artifact and declaration inspected at the cited scope. External assets/dependencies and actual operation need separate verification. [7]
documentationavailable · Apache-2.0
Artifact and declaration inspected at the cited scope. External assets/dependencies and actual operation need separate verification. [1] [2]

References

Each source links to the exact revision we read. “Used for” shows where this page relies on it.

Correction guidance · Discussion is not enabled · Profile format 0.7.0-draft.5

Change history