{"schema":"postcutoff/model@1","as_of":"2026-11-13T13:49:00+01:00","url":"https://postcutoff.com/m/molmoact-2/","md":"https://postcutoff.com/m/molmoact-2/index.md","disclosure":{"written_by":"AI agents (Claude Opus 5.5 in Claude Code)","editor":"human editor","policy":"https://postcutoff.com/about/"},"license":null,"id":"molmoact-2","name":"MolmoAct 2 / MolmoAct 2-Think","org":"Ai2 (Allen Institute for AI)","family":"MolmoAct","released":"2026-05-05","status":"current","type":"robotics","modality_in":["text","image"],"modality_out":["action"],"open_weights":true,"model_license":"Apache-2.0 (code); model weights on HF (license tag not stated on card)","context_window":null,"max_output":null,"knowledge_cutoff":null,"pricing":null,"price_line":"Price not published","access":[{"provider":"Hugging Face","model_id":"allenai/MolmoAct2","url":"https://huggingface.co/collections/allenai/molmoact2-models"},{"provider":"GitHub","url":"https://github.com/allenai/molmoact2"},{"provider":"Hugging Face LeRobot","model_id":"allenai/MolmoAct2-LIBERO-LeRobot","url":"https://huggingface.co/allenai/MolmoAct2-LIBERO-LeRobot"}],"capabilities":[{"name":"Open action reasoning model","detail":"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).","first":false,"discovered":"launch","source":"https://allenai.org/blog/molmoact2"},{"name":"Strong out-of-the-box real-world success","detail":"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).","first":false,"discovered":"launch","source":"https://allenai.org/blog/molmoact2"},{"name":"Fast inference","detail":"~180 ms per action call (790 ms with adaptive depth reasoning) vs ~6,700 ms for the original MolmoAct (up to 37x faster).","first":false,"discovered":"launch","source":"https://allenai.org/blog/molmoact2"},{"name":"Largest open bimanual dataset","detail":"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.","first":false,"discovered":"launch","source":"https://allenai.org/blog/molmoact2"}],"entry":"2026-05-05-ai2-molmoact-2","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.","verified":"2026-09-29","body_md":"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.\n\nSources: [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).","page_url":"https://postcutoff.com/m/molmoact-2/","standing":{"kind":"robotics","newer_count":11,"newer":[{"id":"praxis-1","name":"Praxis-1","org":"Runway","released":"2026-09-30","page_url":"https://postcutoff.com/m/praxis-1/"},{"id":"figure-helix-2-5","name":"Helix 2.5","org":"Figure AI","released":"2026-09-17","page_url":"https://postcutoff.com/m/figure-helix-2-5/"},{"id":"unifolm-wla-1-0","name":"UnifoLM-WLA-1.0","org":"Unitree Robotics","released":"2026-09-10","page_url":"https://postcutoff.com/m/unifolm-wla-1-0/"},{"id":"skild-s1","name":"Skild S1 (Skild Brain)","org":"Skild AI","released":"2026-08-25","page_url":"https://postcutoff.com/m/skild-s1/"},{"id":"generalist-gen-1-5","name":"Generalist GEN-1.5","org":"Generalist AI","released":"2026-08-19","page_url":"https://postcutoff.com/m/generalist-gen-1-5/"}],"before":null,"after":null,"weights":"public","licence":"Apache-2.0 (code); model weights on HF (license tag not stated on card)","licence_permissive":false,"summary":"Weights are public. Licence: Apache-2.0; model weights on HF. See the licence for commercial terms."}}