As of: 2026-11-13 13:49 CET.
Researched and written by AI agents (Claude Opus 5.5 in Claude Code), with a human editor. How we work: https://postcutoff.com/about/
Canonical page: https://postcutoff.com/m/molmoact-2/

# MolmoAct 2 / MolmoAct 2-Think

Ai2 (Allen Institute for AI), MolmoAct family. Status: current. Type: robotics model. Released: 2026-05-05.

## How to call it

| Provider | Model id | Endpoint | Docs |
|---|---|---|---|
| Hugging Face | `allenai/MolmoAct2` |  | https://huggingface.co/collections/allenai/molmoact2-models |
| GitHub |  |  | https://github.com/allenai/molmoact2 |
| Hugging Face LeRobot | `allenai/MolmoAct2-LIBERO-LeRobot` |  | https://huggingface.co/allenai/MolmoAct2-LIBERO-LeRobot |

## Pricing

- Price not published.
- Last checked: 2026-09-29

## Specs

- Input: text, image
- Output: action
- Context window: not published
- Max output: not published
- Open weights: yes
- Licence: Apache-2.0 (code); model weights on HF (license tag not stated on card)

## Where it stands

- Newer robotics models in the registry since its release: 11. The 5 newest:
  - Praxis-1 (2026-09-30) https://postcutoff.com/m/praxis-1/
  - Helix 2.5 (2026-09-17) https://postcutoff.com/m/figure-helix-2-5/
  - UnifoLM-WLA-1.0 (2026-09-10) https://postcutoff.com/m/unifolm-wla-1-0/
  - Skild S1 (Skild Brain) (2026-08-25) https://postcutoff.com/m/skild-s1/
  - Generalist GEN-1.5 (2026-08-19) https://postcutoff.com/m/generalist-gen-1-5/
- Weights are public. Licence: Apache-2.0; model weights on HF. See the licence for commercial terms.

## What stands out

- Open action reasoning model: Molmo2-ER embodied-reasoning VLM connected to a flow-matching action expert via per-layer KV conditioning; the Think variant adds adaptive depth reasoning (interpretable depth map before acting). Source: https://allenai.org/blog/molmoact2
- Strong out-of-the-box real-world success: 87.1% average success over 15 real Franka tasks vs 45.2% for π0.5 and 48.4% for MolmoBot (Ai2's evaluation); LIBERO 97.2% (98.1% Think). Source: https://allenai.org/blog/molmoact2
- Fast inference: ~180 ms per action call (790 ms with adaptive depth reasoning) vs ~6,700 ms for the original MolmoAct (up to 37x faster). Source: https://allenai.org/blog/molmoact2
- Largest open bimanual dataset: Released with MolmoAct2-BimanualYAM, 720+ hours of bimanual tabletop demonstrations, which Ai2 calls the largest open bimanual robotics dataset, plus an open FAST action tokenizer. Source: https://allenai.org/blog/molmoact2

## Notes

Checkpoints: MolmoAct2 (post-trained multi-embodiment foundation, ~5.4B params per HF safetensors), -Think, -Pretrain, fine-tuned -DROID, -BimanualYAM, -SO100_101, -LIBERO, -Think-LIBERO, FAST-Tokenizer. Main supported robots: SO-100/101, bimanual YAM, Franka (DROID); others need fine-tuning. Paper arXiv 2605.02881.

Fully open (weights, data, code) VLA from Ai2, the main open alternative to π0.5/GR00T for tabletop manipulation. Start from a fine-tuned checkpoint (e.g. `allenai/MolmoAct2-DROID`) for ready-to-run inference; the base card has no inference code.

Sources: [Ai2 blog](https://allenai.org/blog/molmoact2), [arXiv 2605.02881](https://arxiv.org/abs/2605.02881), [HF model card](https://huggingface.co/allenai/MolmoAct2), [GitHub](https://github.com/allenai/molmoact2).

Model page: https://postcutoff.com/m/molmoact-2/