MolmoAct2: Action Reasoning Models for Real-world Deployment

2026-05-04

AI Summary

This work addresses the limitations of existing vision–language–action (VLA) models in real-world robotic deployment—namely their closed nature, high cost, high latency, and low success rates—by introducing MolmoAct2, an open-source VLA model featuring three key innovations: a dedicated vision–language backbone (MolmoER), an open action tokenizer (OpenFAST) paired with a flow-matching continuous-action expert, and an adaptive deep reasoning mechanism (MolmoThink). Trained on a large-scale dual-arm robot dataset using a “specialize-then-rehearse” strategy, MolmoAct2 outperforms strong baselines such as Pi-05 across seven simulation and real-world benchmarks and surpasses GPT-5 and Gemini Robotics ER-1.5 on 13 embodied reasoning tasks. The model, along with its code and data, is fully open-sourced.

Abstract

Vision-Language-Action (VLA) models aim to provide a single generalist controller for robots, but today's systems fall short on the criteria that matter for real-world deployment. Frontier models are closed, open-weight alternatives are tied to expensive hardware, reasoning-augmented policies pay prohibitive latency for their grounding, and fine-tuned success rates remain below the threshold for dependable use. We present MolmoAct2, a fully open action reasoning model built for practical deployment, advancing its predecessor along five axes. We introduce MolmoER, a VLM backbone specialized for spatial and embodied reasoning, trained on a 3.3M-sample corpus with a specialize-then-rehearse recipe. We release three new datasets spanning low-to-medium cost platforms, including MolmoAct2-BimanualYAM, 720 hours of teleoperated bimanual trajectories that constitute the largest open bimanual dataset to date, together with quality-filtered Franka (DROID) and SO100/101 subsets. We provide OpenFAST, an open-weight, open-data action tokenizer trained on millions of trajectories across five embodiments. We redesign the architecture to graft a flow-matching continuous-action expert onto a discrete-token VLM via per-layer KV-cache conditioning. Finally, we propose MolmoThink, an adaptive-depth reasoning variant that re-predicts depth tokens only for scene regions that change between timesteps, retaining geometric grounding at a fraction of prior latency. In the most extensive empirical study of any open VLA to date, spanning 7 simulation and real-world benchmarks, MolmoAct2 outperforms strong baselines including Pi-05, while MolmoER surpasses GPT-5 and Gemini Robotics ER-1.5 across 13 embodied-reasoning benchmarks. We release model weights, training code, and complete training data. Project page: https://allenai.org/blog/molmoact2

Problem

• Vision-Language-Action
• real-world deployment
• action reasoning
• embodied reasoning
• open-weight models

Innovation

• MolmoAct2
• embodied reasoning
• open-weight VLA
• adaptive-depth reasoning
• continuous-action expert