Big techChinaFounded 2017
Alibaba Group research institute that publishes the open Rynn series of embodied models (RynnVLA, RynnEC, RynnBrain) and WorldVLA.
Alibaba DAMO Academy is the research arm of Alibaba Group, announced on 2017-10-11 as a global research program with seven initial labs in Beijing, Hangzhou, San Mateo, Bellevue, Moscow, Tel Aviv and Singapore. Its robotics work, grouped as the RynnBot family, includes RynnVLA-001, a vision-language-action model built on a pretrained video generation model and released in August 2025.
Sources: alibabagroup.com (11 Oct 2017) · github.com (17 Oct 2026)
Corporate division · Research division of Alibaba Group; no separate funding
#2 of 104 model companies · score 91.3 / 100 · updated 19 Oct 2026
Each bar shows an input from 0 to 100% before weighting. 6 of 6 listed models have open weights. Most-starred official repository: alibaba-damo-academy/RynnVLA-002 (1.1k 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 weightsEmbodied reasoning
Second release of the RynnBrain embodied foundation models, with real-robot VLA evaluation on Unitree G1, Astribot and Tianji-Wuji platforms.
No benchmark results on a shared public benchmark have been published for this model.
Open weightsEmbodied reasoning
Open spatiotemporal embodied foundation model family built on Qwen3-VL, with post-trained Nav, Plan, VLA and CoP variants.
No benchmark results on a shared public benchmark have been published for this model.
Open weightsVLA
Successor to WorldVLA that combines a VLA policy and a world model in one framework, adding a continuous action transformer, wrist-camera input and state input.
| Benchmark | Result | Source |
|---|---|---|
| LIBEROAverage success rate across LIBERO-Spatial/Object/Goal/Long (%) | 97.4% | arxiv.orgTable 1, RynnVLA-002-Continuous, no pretraining (Spatial 99.0 / Object 99.8 / Goal 96.4 / Long 94.4); discrete-action variant averages 93.3. As of 21 Nov 2025 |
Open weightsVLA
VLA pretrained with ego-centric video generation on 12M human manipulation videos, then trajectory and action fine-tuning; evaluated on LeRobot SO100 arms.
No benchmark results on a shared public benchmark have been published for this model.
Open weightsEmbodied reasoning
Video multimodal LLM for embodied cognition (object and spatial understanding in egocentric video), released with the RynnEC-Bench benchmark.
No benchmark results on a shared public benchmark have been published for this model.
Open weightsWorld model
Autoregressive action world model initialized from Chameleon that unifies action prediction and future-image generation in one model.
