Robots gain upgraded cognition and reach 90% accuracy with Gemini
OrionStar expands global multilingual interaction with Text-to-Speech and Translation API
Saves costs and speeds up iteration with Gemini
OrionStar actively powers personalized, global services and eldercare with Vertex AI
By integrating Gemini, OrionStar strengthens AgentOS with multimodal cognition and interaction, reshaping human-robot collaboration, boosting efficiency across scenarios, and accelerating global expansion.
By integrating Gemini, OrionStar strengthens AgentOS with multimodal cognition and interaction, reshaping human-robot collaboration, boosting efficiency across scenarios, and accelerating global expansion.
Gemini-powered acceleration: Scaling OrionStar’s interactive robots globally
In the early stage of the robotics industry, robots largely relied on preset commands and passive responses, lacking true understanding and contextual awareness. Users often had to wake them multiple times just to complete a simple conversation, making “intelligence” feel mechanical and disconnected.
Founded in 2016, OrionStar was born to build “truly useful robots.” Today, its business spans over 60 countries and regions, serving more than 20,000 customers across 20+ application scenarios, including exhibitions, elderly care, education and healthcare.
Under the leadership of Chairman Fu Sheng, OrionStar built its proprietary robot operating system, AgentOS, as a multi-model cognitive hub that enables coordinated intelligence. This allows robots to move beyond passive command execution toward intent- and context-aware understanding and decision-making. As part of its global expansion, OrionStar partnered with Google Cloud to integrate Gemini into its multimodal interaction pipeline, working in close synergy with AgentOS.
With Gemini multimodal capabilities, speech, vision, and motion decision-making are deeply integrated. Robots can now recognize lip movements, interpret visual information, and engage in natural multi-turn conversations — all without needing a specific activation phrase (like “Hey”). Since the integration, interaction accuracy has reached 90%, delivering a more natural and coherent conversational experience — allowing users to feel that “robots speak like humans.”
Powered by Gemini, we’re accelerating our mission to enable robots to think, act, and communicate like humans. With OrionStar’s AgentOS, Gemini Live’s low-latency capabilities enable smoother multi-turn dialogue and interaction, seamlessly integrating speech, vision, and action—allowing conversations with robots to truly “flow."
Fu Sheng
Chairman, OrionStar
OrionStar AgentOS: Building an efficient and scalable robot operating system
The launch of AgentOS marks OrionStar’s entry into a new stage of intelligence, positioning it as an industry leader with a self-developed multi-agent robotic operating system. By integrating Gemini into its multimodal pipeline, OrionStar further enhances robots’ understanding and orchestration capabilities—making robots smarter while enabling more efficient and systematized enterprise operations.
We have upgraded the “Large Model Robot Formula.” Only when advanced AI technology, AgentOS, reliable hardware, and comprehensive services come together can robots truly create value. As Gemini is deeply integrated into OrionStar’s AgentOS, we have achieved a new level of global expansion, empowering robots to realize their true potential and deliver real value in the world.
Fu Sheng
Chairman, OrionStar
OrionStar AgentOS unifies speech, vision, cognition, and decision-making modules, enabling multi-agent collaboration and intelligent orchestration. Robots gain team-like intelligence, allowing them to coordinate, adapt, and learn collectively — much like humans working together toward shared goals.
On the development side, with Gemini 2.5 Pro for AI coding capability, OrionStar efficiently manages sub-agents, significantly reducing R&D costs, accelerating product iteration, and boosting operational efficiency across teams.
In 2024, Chairman Fu Sheng introduced the upgraded “Iron Quadrathlon” framework, which combines AI technology, AgentOS, reliable hardware, and LLM robust services as the foundation for creating sustainable value in robotics.
In synergy with OrionStar’s AgentOS, Gemini’s powerful multimodal cognition not only gives robots a “thinking brain,” but also equips enterprises with greater agility, scalability, and operational efficiency.
Accelerating the global expansion of voice-interactive robots, evolving the future of human-robot collaboration
As OrionStar expands globally, multilingual communication and localized deployment have become major challenges. Each country has its own languages, cultures, and operational standards, yet robot services must deliver a consistent, stable, and natural interaction experience.
To address these challenges, OrionStar introduced Google Cloud’s models and capabilities into its existing technology stack to support the coordination of speech, vision, and knowledge orchestration, explore fluent multilingual interaction, and evaluate localized deployment strategies. By integrating Google Cloud’s Gemini models, OrionStar demonstrated consistent and reliable intelligent interaction during a demo at a Singapore exhibition, with plans to gradually expand to more countries and regions as the business progresses.
Today, OrionStar’s robots are widely deployed across Europe, Japan, South Korea, and North America, serving diverse scenarios such as exhibitions, elderly care, education, and healthcare.
Looking ahead, we aim to bring truly useful robots to more scenarios, enable wider adoption at scale, and deliver more globalized services. By letting robots handle the trivial and repetitive tasks, we empower people to focus on innovation and to create things of real value.
Fu Sheng
Chairman, OrionStar
For Europe’s highly complex multilingual environment, the Gemini-powered robot demo showcased in Singapore provided early validation of Gemini’s potential in language understanding and real-time response, positioning it as a promising option for future global voice interaction. In parallel, OrionStar is also exploring the use of Vertex AI in eldercare scenarios to enable companionship services with emotional expression and personalized experiences.
By leveraging the AI ecosystem in Google Cloud, OrionStar is shaping robots that take on repetitive tasks, allowing humans to focus on creativity and exploration, redefining the future of intelligent collaboration between humans and robots.
OrionStar, founded in 2016, was born to build "truly useful" robots. It combines AI and robotics to serve 20,000+ customers in 60 countries across 20+ scenarios, including exhibitions, eldercare, education, healthcare, and the Beijing 2022 Winter Games.
Industry: Manufacturing
Location: China
Products: Google Cloud, Gemini, Cloud Translation API, Text-to-Speech AI, Vertex AI
