Research labSeattle, WA, United StatesFounded 2014
Non-profit AI research institute that releases the open MolmoAct family of action reasoning models for robot manipulation.
Ai2 is a Seattle-based non-profit AI research institute founded in 2014 by Paul Allen. It releases open models, data and code, including the Molmo vision-language models. In August 2025 it released MolmoAct, an open 7B action reasoning model, and in May 2026 MolmoAct2, which adds a flow-matching action expert and released weights, training code and datasets.
Sources: allenai.org (19 Oct 2026) · allenai.org (12 Aug 2025) · huggingface.co (19 Oct 2026)
Research lab · Non-profit research institute; models released openly, no venture funding.
#10 of 104 model companies · score 78.0 / 100 · updated 19 Oct 2026
Each bar shows an input from 0 to 100% before weighting. 2 of 2 listed models have open weights. Most-starred official repository: allenai/molmoact2 (813 stars, 19 Oct 2026).
The score (0–100) measures public, checkable evidence about a company's robot models. It does not measure private capability. A company that publishes no benchmark results and no weights can still have strong models. Every input is computed from the data on each company page, and every number there links to its source.
| Input | Weight | How it is measured |
|---|---|---|
| Published benchmarks | 35% | For each scored benchmark, the company's best result divided by the best result any company has published, averaged over the benchmarks it reports. No published results scores 0. Results come only from the model's own paper, model card or the official benchmark leaderboard. Scored: LIBERO (4-suite average success, %); SimplerEnv Google Robot, visual matching (%); SimplerEnv Google Robot, variant aggregation (%); SimplerEnv WidowX / Bridge (%); CALVIN ABC→D (average length, 0–5); RoboArena (pairwise rating). |
| Open weights | 20% | Share of the company's listed models whose weights the public can download. |
| Adoption | 20% | GitHub stars on the company's most-starred official model repository, on a log scale relative to the most-starred repository in the list. Hugging Face downloads are shown on company pages but not scored, because gated repositories report zero. |
| Released models | 15% | Number of publicly announced, named models, capped at 5. |
| Disclosed funding | 10% | The sum of the funding rounds listed on the company's profile that were disclosed in USD, on a log scale from $10M to the largest sum in the list. Rounds reported only in other currencies are listed but not converted or counted. This input applies only to private startups and robot makers. For big-tech divisions, research labs and listed companies it is n/a, and its weight is shared across the other inputs in proportion. |
Sources. arXiv papers and official model cards for benchmarks and parameters; official leaderboards (RoboArena); the GitHub and Hugging Face APIs for stars and downloads; company announcements and established press for funding. Missing values are shown as n/a, never estimated.
Updated 19 Oct 2026. Benchmark protocols differ between papers, for example in fine-tuning and number of trials, so small gaps between close scores are not meaningful.
Open weightsVLA
Successor to MolmoAct built on the Molmo2-ER embodied-reasoning VLM (4B LLM) with a flow-matching continuous action expert conditioned on per-layer VLM keys and values; checkpoints released for bimanual YAM, SO-100/101 and DROID Franka.
Open weightsVLA
Action reasoning model that predicts depth perception tokens and an image-space visual reasoning trace before emitting action tokens, built on the Molmo VLM.
| Benchmark | Result | Source |
|---|---|---|
| LIBEROAverage success rate across LIBERO-Spatial/Object/Goal/Long (%) | 86.6% | arxiv.orgTable 2 caption: MolmoAct average success rate across the four LIBERO suites (fine-tuned MolmoAct-7B-D). As of 18 Sep 2025 |
| SimplerEnv Google VMSimplerEnv Google Robot visual matching average (%) | 71.6% | arxiv.orgTable 1 and Sec. 5 text: MolmoAct-7B-D fine-tuned on RT-1 data; zero-shot pretrain checkpoint reported 70.5%. As of 18 Sep 2025 |
| SimplerEnv Google VASimplerEnv Google Robot variant aggregation average (%) | 72.1% | arxiv.orgTable 1, MolmoAct (fine-tuned) row, variant aggregation overall column (column assignment read from HTML table layout). As of 18 Sep 2025 |
