# Pudu Robotics Showcases Embodied AI and New Models at BEYOND Expo

> Source: <https://letsdatascience.com/news/pudu-robotics-showcases-embodied-ai-and-new-models-at-beyond-52179713>
> Published: 2026-06-03 11:22:19.824719+00:00

# Pudu Robotics Showcases Embodied AI and New Models at BEYOND Expo

Pudu Robotics exhibited its latest commercial service robots and platform work at **BEYOND Expo 2026** in Macao and Founder and CEO **Felix Zhang** spoke on the opening ceremony, according to PRNewswire and the company website. Pudu and event organizers reported deployments of service robots in over **80 countries and regions**, with company materials citing **120,000** robots in operation and PRNewswire/ANTARA noting **130,000** in circulation. Per Pudu Robotics' announcement, the company introduced the embodied intelligence foundation model PuduFM 1.0 and a general embodied AI agent platform, and displayed a multi-form product portfolio including semi-humanoid, quadruped, and a full-size biped humanoid robot. Industry context: observers see this as another data- and platform-driven push to move robotics from single-task automation toward multi-form, embodied AI systems.

### What happened

Pudu Robotics exhibited at **BEYOND Expo 2026** in Macao and Founder & CEO **Felix Zhang** participated in a featured fireside chat, per PRNewswire and the company's event announcement. The company and event organizers reported global deployments across more than **80 countries and regions**, with the company website and BEYOND Expo text citing **120,000** robots in service while PRNewswire and ANTARA used a **130,000** figure. Per Pudu Robotics' press materials, the company introduced an embodied intelligence foundation model, PuduFM 1.0, and a general embodied AI agent platform, and showcased multiple robot form factors including semi-humanoid service robots, an industry-grade quadruped, and a full-size biped humanoid.

### Technical details

Per Pudu Robotics' announcement, PuduFM 1.0 is described by the company as focusing on **deep perception and reasoning in 3D space**, prediction of future physical states, and a continuous learning loop described as "learning, anticipating, validating, correcting." The company frames its software architecture as "One Brain, Multiple Embodiments," which it says enables capability transfer across specialized robots, semi-humanoid units, and humanoid platforms, according to PRNewswire and the firm website.

Editorial analysis - technical context: Robotics practitioners adopting embodied-AI approaches typically face three technical challenges: aligning perception stacks with real-world dynamics, scaling sim-to-real or real-world continual learning pipelines, and building transferable control abstractions that work across morphologies. Industry patterns show that foundation-model-style approaches for robots place a premium on large-scale, labeled and unlabeled interaction data plus modular runtime stacks that combine perception, prediction, and motion primitives.

### Context and significance

Industry context: The announcements fit a broader trend where service-robot vendors pair hardware portfolios with centralized model and data platforms to accelerate feature parity across form factors. Recent vendor claims and coverage increasingly emphasise multi-form ecosystems and software-first value capture, reflecting investor and operator preference for recurring software updates and platform services over one-off hardware sales. For operators and integrators, the combination of a foundation model claim plus an agent platform signals vendor focus on delivering higher-level cognition and autonomy features, which can affect integration effort and lifecycle upgrade planning.

### What to watch

Observers following the sector will watch for three practical indicators: whether Pudu publishes developer SDKs or APIs that expose PuduFM capabilities to integrators; independent field benchmarks or third-party deployments that validate the claimed 3D reasoning and prediction improvements; and data-sharing or fleet-management partnerships that would materially expand the dataset behind the foundation model. Additional signals include commercial availability and pricing of the full-size biped humanoid and the quadruped, plus any safety, compliance, or operations case studies from early customers.

Editorial analysis: For practitioners, the most immediate questions are operational. Companies deploying service robots typically evaluate robustness in constrained sites, integration complexity with existing IT and teleoperation tooling, and the vendor's capacity to deliver over-the-air updates without disrupting service. As embodied-AI claims become routine in vendor marketing, independent testing and transparent benchmarks will grow more valuable for procurement teams evaluating multi-form robot platforms.

## Scoring Rationale

The story is a notable product and platform update from a leading service-robotics vendor that matters to operators and integrators. It is not a frontier-model breakthrough, but the introduction of an embodied foundation model plus multi-form hardware affects procurement, integration, and field validation decisions.

Practice interview problems based on real data

1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.

[Try 250 free problems](/problems)
