MolmoAct2-DROID

MolmoAct2 is an open vision-language-action model for robot control. It builds on Molmo2-ER and attaches a flow-matching continuous action expert that conditions on the VLM key-value cache through a per-layer connection.

This checkpoint is fine-tuned on the filtered DROID Franka mixture with absolute joint-pose control. It is intended for both further fine-tuning and DROID-style policy inference.

Quick Links

• 📂 Models: Models, Finetuned Models
• 📂 Datasets: MolmoAct2-BimanualYAM Dataset, MolmoAct2 Datasets, Molmo2-ER Datasets
• 📄 Paper: arXiv:2605.02881
• 💻 Code: allenai/molmoact2
• 🎥 Blog Post: MolmoAct2

Intended Use

Use this checkpoint for DROID-style inference or for further fine-tuning. Dataset normalization metadata is stored in norm_stats.json. pass norm_tag="franka_droid" at inference time.

Continuous action prediction is the intended and recommended inference mode. Discrete action prediction is exposed for parity and debugging, but we use continuous actions by default.

Install

pip install torch transformers pillow numpy huggingface_hub

Sample Input

This sample comes from allenai/droid_lerobot, episode 1, frame 0. The example uses the exterior 1 camera, a copied exterior 1 view, and then the wrist camera as input.

Exterior 1 RGB	Copied Exterior 1 RGB	Wrist RGB

from huggingface_hub import hf_hub_download
from PIL import Image
import numpy as np

repo_id = "allenai/MolmoAct2-DROID"

exterior_1_rgb = Image.open(
    hf_hub_download(repo_id, "assets/sample_exterior_1_left_rgb.png")
).convert("RGB")
wrist_rgb = Image.open(
    hf_hub_download(repo_id, "assets/sample_wrist_left_rgb.png")
).convert("RGB")

task = "Put the black objects into the drawer and close the drawer."
robot_state = np.array(
    [
        -0.12726949,
        -0.30641943,
        0.09134164,
        -2.4143615,
        -0.26460838,
        2.068765,
        0.123698,
        0.0,
    ],
    dtype=np.float32,
)

Continuous Actions

import numpy as np
import torch
from huggingface_hub import hf_hub_download
from PIL import Image
from transformers import AutoModelForImageTextToText, AutoProcessor

repo_id = "allenai/MolmoAct2-DROID"

exterior_1_rgb = Image.open(
    hf_hub_download(repo_id, "assets/sample_exterior_1_left_rgb.png")
).convert("RGB")
wrist_rgb = Image.open(
    hf_hub_download(repo_id, "assets/sample_wrist_left_rgb.png")
).convert("RGB")
task = "Put the black objects into the drawer and close the drawer."
robot_state = np.array(
    [
        -0.12726949,
        -0.30641943,
        0.09134164,
        -2.4143615,
        -0.26460838,
        2.068765,
        0.123698,
        0.0,
    ],
    dtype=np.float32,
)

processor = AutoProcessor.from_pretrained(repo_id, trust_remote_code=True)
model = AutoModelForImageTextToText.from_pretrained(
    repo_id,
    trust_remote_code=True,
    dtype=torch.float32,
).to("cuda").eval()

out = model.predict_action(
    processor=processor,
    images=[exterior_1_rgb, exterior_1_rgb, wrist_rgb],
    task=task,
    state=robot_state,
    norm_tag="franka_droid",
    inference_action_mode="continuous",
    enable_depth_reasoning=False,
    num_steps=10,
    normalize_language=True,
    enable_cuda_graph=True,
)

actions = out.actions

MolmoAct2 was trained with mixed precision. For our reported experiments, we ran inference in float32. This path uses the most GPU memory: roughly 26GB with CUDA graph enabled, or around 24GB without CUDA graph.

If you have a GPU with less memory, you can run inference with bfloat16 instead:

model = AutoModelForImageTextToText.from_pretrained(
    repo_id,
    trust_remote_code=True,
    dtype=torch.bfloat16,
).to("cuda").eval()

with torch.inference_mode(), torch.autocast("cuda", dtype=torch.bfloat16):
    out = model.predict_action(...)

Using bfloat16 is much more memory efficient and can run under 16GB of GPU memory in our tests. It usually does not hurt performance much.

images should preserve camera order. The DROID checkpoint was fine-tuned with three camera slots ordered as exterior 1, exterior 2, and wrist. In DROID-style inference setups that expose only one exterior camera and one wrist camera, duplicate the exterior image and pass [exterior_1_rgb, exterior_1_rgb, wrist_rgb]. This copied-exterior input matches the training camera layout more closely and might improve performance. However, two-camera inputs such as [exterior_1_rgb, wrist_rgb] are also acceptable and should also work. Images may be PIL images or RGB arrays. state is the raw robot state, and actions are returned in robot scale.

normalize_language=True is the default. It lowercases the task string and removes trailing sentence punctuation to match training preprocessing. Set it to False if you need to preserve the task text exactly.

enable_cuda_graph=True is the default. The first few calls can be slow because the model warms up and captures CUDA graphs. run several random warm-up calls before measuring deployment latency. num_steps controls the continuous flow solver and defaults to the checkpoint config value, 10.

Depth reasoning is disabled for this checkpoint. Calling enable_depth_reasoning=True will raise an error.

Discrete Actions

Discrete action inference requires a caller-provided action tokenizer. It is not saved in this repository. Discrete mode decodes action tokens directly. the continuous action expert is not used.

action_tokenizer = AutoProcessor.from_pretrained(
    "allenai/MolmoAct2-FAST-Tokenizer",
    trust_remote_code=True,
)

out = model.predict_action(
    processor=processor,
    images=[exterior_1_rgb, exterior_1_rgb, wrist_rgb],
    task=task,
    state=robot_state,
    norm_tag="franka_droid",
    inference_action_mode="discrete",
    action_tokenizer=action_tokenizer,
    enable_depth_reasoning=False,
)

Model and Hardware Safety

MolmoAct2 generate robot actions from visual observations and language instructions, but their behavior may vary across embodiments, environments, and hardware configurations. Users should carefully validate model outputs before deployment, especially when operating physical robots or other actuated systems. Where possible, actions should be monitored through interpretable intermediate outputs (adaptive depth map), simulation rollouts, action limits, or other safety checks before execution on hardware. The model’s action space should be bounded by the training data, robot controller limits, and task-specific safety constraints, including limits on speed, workspace, torque, and contact force. Users should follow the hardware manufacturer’s safety guidelines, use appropriate emergency-stop mechanisms, and operate the system only in a safely configured environment with human supervision.

Citation

@misc{fang2026molmoact2actionreasoningmodels,
      title={MolmoAct2: Action Reasoning Models for Real-world Deployment}, 
      author={Haoquan Fang and Jiafei Duan and Donovan Clay and Sam Wang and Shuo Liu and Weikai Huang and Xiang Fan and Wei-Chuan Tsai and Shirui Chen and Yi Ru Wang and Shanli Xing and Jaemin Cho and Jae Sung Park and Ainaz Eftekhar and Peter Sushko and Karen Farley and Angad Wadhwa and Cole Harrison and Winson Han and Ying-Chun Lee and Eli VanderBilt and Rose Hendrix and Suveen Ellawela and Lucas Ngoo and Joyce Chai and Zhongzheng Ren and Ali Farhadi and Dieter Fox and Ranjay Krishna},
      year={2026},
      eprint={2605.02881},
      archivePrefix={arXiv},
      primaryClass={cs.RO},
      url={https://arxiv.org/abs/2605.02881}, 
}