Explore the Ecosystem

Alphabet-owned AI robotics company building software and AI for industrial robotics; developer and open-source publisher of Intrinsic Core, an open runtime and real-time control framework released at ROSCon 2026.

NVIDIA is the world's leading AI computing company, providing GPU-accelerated hardware and software platforms for AI, robotics, and simulation, including the Isaac robotics platform and Jetson edge AI hardware.

Hugging Face is the company behind the Hugging Face Hub, an open platform for hosting and sharing AI models, datasets and applications, and the organization behind the open-source robotics project LeRobot.

World Labs is a spatial intelligence company building frontier world models that perceive, generate, reason, and interact with virtual and physical worlds.

Pollen Robotics is a Bordeaux-based open-source robotics company founded in 2016, maker of the Reachy humanoid line and the Microduck reinforcement-learning biped; it is part of Hugging Face.

Zeroth Robotics is the robotics company behind the Zeroth Bridge small humanoid robot and the OpenBridge open-source ecosystem; its product lines include the M1 companion robot, the W1 outdoor robot, and the Jupiter humanoid robot.

Beijing Humanoid Robot Innovation Center is a Beijing-based full-stack embodied intelligence technology company that operates the National and Local Co-built Embodied Intelligence Robot Innovation Center and develops the Tiangong humanoid robot family and the Pelican model line.

Leju Robotics is a Chinese humanoid robot company and the developer of the KUAVO humanoid robot series. It describes itself as a leader in the industrialization of humanoid robots.

Lightwheel is a Physical AI infrastructure company building SimReady assets, egocentric human data, and evaluation platforms that power robots to learn and deploy at scale.

A robotics data company providing multimodal data acquisition, high-precision motion capture, and teleoperation systems for embodied AI. Its ModalityNet platform and World Compiler approach produce physical AI training data, including the publicly released HiPHI dataset.

AI company developing generative AI models and native omni-modal world models, including the HiDream-O1 model family.

Embodied AI company developing commercial-grade robots and world-action foundation models for real-world operations.

Digit 5 is Agility Robotics' fifth-generation general-purpose humanoid robot, engineered for cooperatively safe work in close proximity to people without physical safety barriers, unveiled September 15, 2026.

Microduck is a 25 cm open-source biped robot from Pollen Robotics (Hugging Face) that moves using reinforcement-learning policies, with US$399 pre-orders opened on August 27, 2026.

Zeroth Bridge is a compact small humanoid robot developed by Zeroth Robotics as the robot body of the OpenBridge open-source ecosystem, with global pre-orders opened on September 2, 2026.

Tiangong Ultra is a track-and-field-oriented humanoid robot in the Tiangong family developed by Beijing Humanoid Robot Innovation Center. At the 2nd World Humanoid Robot Games it set a 100m sprint record of 8.86 seconds.

Tiangong Omni is a lightweight humanoid robot in the Tiangong family, released by Beijing Humanoid Robot Innovation Center at WRC 2026. It won the 400m race (45.66s) and the 400m obstacle race at the 2nd World Humanoid Robot Games.

A high-dynamic humanoid robot prototype unveiled by Unitree Robotics in August 2026 as a research and technology-verification platform. Unitree reports a top speed of 12.66 m/s and a 2.0 m standing high jump; the figures are company-reported and not independently verified.

Boston Dynamics' quadruped mobile robot, officially positioned as an agile mobile robot for dynamic sensing and industrial inspection, with SDK support for custom development.

Apptronik's second-generation full-size humanoid robot, emphasizing human-robot interaction and whole-body execution; the primary demonstration embodiment for Gemini Robotics 2.

Advanced dynamic humanoid from MagicLab (1.8m, 70kg, 31 DOF, 450 Nm torque) capable of autonomous basketball dunking, showcased at WAIC 2026.

Compact desktop-sized emotional companion robot from Fourier Intelligence (~$140), unveiled at WAIC 2026.

Advanced 20-DOF dexterous robotic hand from AGIBOT with vision-based tactile sensing, demonstrated at WAIC 2026.

Transformable humanoid and quadruped robot from Unitree Robotics, capable of carrying a human operator.

AGIBOT's embodied-native omni-modal model for real-world human-robot interaction, built on a Thinker-Talker-Actor architecture that aligns speech, movement, and expression on the same timeline.

X-Planner-9B-0916 is X Square Robot's Qwen3.5-based 9B multimodal planning front end for long-horizon robot manipulation, released September 16, 2026 with Apache 2.0 open weights.

UnifoLM-ER-Flow is Unitree's interaction-centric world-model extension of UnifoLM-ER-1, introducing discrete action learning and released as open weights on September 11, 2026 under the UnifoLM-WLA-1.0 plan.

UnifoLM-ER-1 is Unitree's 4B-parameter vision-language-action foundation model, open-sourced on September 11, 2026 under the UnifoLM-WLA-1.0 open-source plan and built on a Qwen3-VL-4B backbone.

GE-Act 2.0 (Genie Envisioner Act 2.0) is a world-action model for robotic manipulation whose generative and action components are pretrained from scratch on manipulation data.

Atlas is World Labs' next-generation omni world model, pretrained from scratch to natively operate on text, images, video, and 3D.

Pelican-Unify 1.0 is an embodied-intelligence unified foundation model developed by Beijing Humanoid Robot Innovation Center that unifies perception, language, reasoning, and action control in a single neural network.

LET-KUAVO-VLA-1.0 is an open-source vision-language-action (VLA) model released by Leju Robotics for its KUAVO humanoid robot platform. The gated Hugging Face release provides model weights for the KUAVO VLA stack.

GEAR-SONIC (SONIC) is a humanoid behavior foundation model developed by NVIDIA. It learns a unified whole-body control policy from large-scale human motion data, enabling walking, running, crawling, jumping and manipulation.

A vision-language-action model for surgical robotics developed by NVIDIA. It is a post-trained variant of the NVIDIA Isaac GR00T N1.6 foundation, adapted for surgical robots using the Open-H embodiment dataset.

A hierarchical robot foundation model with world-model-guided test-time computation, developed by Shanghai Innovation Institute. The low-level vision-language-action policy is released as an Apache-2.0 pretrained checkpoint built on a Qwen3.5-2B backbone.

An interactive world model developed by HiDream.ai, built on its native omni-modal UiT architecture.

AGIBOT WORLD 2026 Theme 3 is AGIBOT's open-source real-world robot reinforcement learning dataset, whose first release includes 11,430 trajectories spanning expert demonstrations, autonomous policy rollouts and human-in-the-loop corrections.

A staged 100,000-hour open release of egocentric human data for Physical AI — 15,000+ tasks across 18 task categories, 15,000+ scenes and 7 environments, with hand/body pose and event-level annotation, published by Lightwheel in partnership with Hugging Face.

A high-precision optical motion-capture dataset released by Noitom Robotics at the World Robot Conference in August 2026. It contains 617.5 hours of whole-body human motion and human-object interaction from 132 performers, captured at 90 Hz with sub-millimeter marker tracking. Noitom describes it as among the largest high-precision human-motion datasets ever made publicly available.

A benchmark dataset for capturing hand grasping of objects, providing 2D/3D annotations for hand pose estimation, 6D object pose estimation, and robot-to-human object handover tasks.

A richly-annotated, large-scale repository of 3D CAD models organized under the WordNet taxonomy, containing millions of models across thousands of semantic categories with consistent rigid alignments, parts, and symmetry annotations.

A large dataset of high-fidelity 3D objects with 800K+ annotated 3D models across diverse categories, designed to support 3D vision, embodied AI, and simulation research.

A large-scale egocentric video benchmark recorded by 32 participants in their native kitchen environments, capturing 100 hours of non-scripted daily kitchen activities with 90K+ action segments.

A massive-scale egocentric video dataset with 3,670 hours of daily-life activity video spanning hundreds of scenarios, captured by 931 unique camera wearers from 74 worldwide locations across 9 countries.

A framework for procedural generation of Embodied AI environments, enabling sampling of arbitrarily large datasets of diverse, interactive, customizable virtual environments for navigation, interaction, and manipulation tasks.

An artist-authored, annotated, reconfigurable 3D dataset of apartments with articulated objects (cabinets, drawers, etc.) designed for training virtual robots in interactive physics-enabled environments, part of Habitat 2.0.

A dataset of 18 highly photo-realistic 3D indoor scene reconstructions with dense meshes, HDR textures, semantic class and instance information, and planar mirror/glass reflectors for embodied AI research.

An RGB-D video dataset containing 2.5 million views across 1,513 indoor scenes annotated with 3D camera poses, surface reconstructions, and semantic segmentations for 3D scene understanding research.

Intrinsic's open-source local runtime, SDK, and hardware-agnostic real-time control framework for industrial robotics, with ROS 2 interoperability and built-in 6-DoF perception, motion planning, and digital twin capabilities.

MuJoCo (Multi-Joint dynamics with Contact) is a general-purpose physics simulator by Google DeepMind with 15,000+ GitHub stars, widely used for robotics and embodied AI research. Version 3.13.0 was released on September 9, 2026.

OpenBridge is the fully open-source (GPLv3) developer ecosystem of Zeroth Robotics, organized into Robot, Skill, and Simulation Hubs around the Zeroth Bridge humanoid robot.

INDEX is Figure's crowdsourced physical-data collection platform and pipeline for training humanoid robots. It pays a global network of contributors to upload real-world task videos, which Figure uses to train its Helix models.

An open standard and ecosystem for robot policy evaluation and deployment that reduces N x M policy-to-environment integrations to O(N+M).

Isaac Lab is an open-source GPU-accelerated robot learning framework from NVIDIA on Isaac Sim, with 30+ environments and 16+ robots for RL, IL, and motion planning.

LeRobot is an open-source robotics library from Hugging Face providing state-of-the-art policies, standardized datasets, and hardware interfaces for democratizing robot learning.

Princeton Robotics project combining LLM-based reasoning with existing robot policies, improving real-robot performance from 16.7% to 97.3%.

REAL is an agentic framework for open-world mobile manipulation, accepted to ECCV 2026, with REAL-Bench (241 tasks) and 78.3% end-to-end success on a physical dual-arm mobile robot.

GigaWorld is an umbrella project developing World Action Models (WAMs) for robot policy learning, including GigaWorld-0, GigaWorld-Policy, GigaWorld-Policy-0.5, and GigaWorld-1.

A family of embodied foundation models for real-world agents, developed by Tencent Robotics X and HY Vision Team. Encompasses embodied VLMs (HY-Embodied-0.5) and VLA policies (HyVLA-0.5).

NVIDIA Isaac GR00T is an open reference platform and development project for general-purpose humanoid robotics, combining open foundation models (GR00T N1 series), data pipelines, simulation frameworks, CUDA-X accelerated runtime libraries, and NVIDIA Jetson Thor for real-time robot inference and control.

A comprehensive multi-turn benchmark for interactive video world model evaluation, developed by the Meituan LongCat team.

Google DeepMind's agentic safety benchmark for embodied agents, evaluating safety orchestration and uncertainty resolution, built on a multimodal image-text-video dataset.

LIBERO (Lifelong Robot Learning Benchmark) is a benchmark suite for studying knowledge transfer in lifelong robot manipulation, featuring 130 tasks across five task suites with standardized evaluation protocols.

A unified sim-and-real benchmark for evaluating generalist robot manipulation policies, with 42 simulation tasks and 18 real-world tasks. Built on Isaac Sim.

BEHAVIOR-1K is a comprehensive embodied AI benchmark with 1,000 everyday activities across 50 scenes and 9,000+ objects, built on the OmniGibson simulation environment for human-centered robotics research.

LIBERO-Pro is a robotic manipulation benchmark designed to evaluate long-horizon task generalization and skill composition.