Built on a 5-nanometer process, the new SoC is designed to support models in the 100 billion parameter class, with full-stack software for private and responsive AI agents at the edge
SINGAPORE, July 23, 2026 /PRNewswire/ — Acrab, a technology company building agentic AI compute infrastructure for the next generation of intelligent systems, today unveiled GΞLIX 1, its first-generation edge AI system-on-chip (SoC), together with Agent Box, a personal edge AI system powered by the company’s full-stack computing platform.
As AI moves from generating answers to completing tasks, agents increasingly need to understand context, remember preferences and coordinate tools and devices in real time. Running these capabilities locally can produce faster responses, keep sensitive information under the user’s control and maintain core functions when cloud connectivity is limited.
For years, models in the 100 billion parameter class have required cloud infrastructure. GΞLIX 1 is designed to bring state-of-the-art AI models at this scale into locally operated edge systems. Powered by GΞLIX, Acrab’s Agent Box is a high-performance personal edge AI center designed to put AI agents into action in a more personal and customized way, with local large model inference, persistent memory, multimodal interactions and agent orchestration capabilities.
By replacing cloud AI’s recurring token-fees per use, Agent Box is a one-time investment with long-term value, hence relieving users’ token anxiety, and allowing AI to move from an occasional tool into an always-available assistant woven into everyday work and life.
“Generative AI helped people find answers. Agentic AI will help them get things done,” said Dr. Ken Phua, CEO of Acrab. “Running models in the 100 billion parameter class on a system small enough to sit on a desk presents a significant computing challenge. GΞLIX 1 is designed to deliver the performance, memory bandwidth and responsive local inference required, while Agent Box shows how that capability can become a complete user experience.”
A private AI center built for everyday life
Agent Box is designed as a private, always-on AI center for personal workspaces and homes. It keeps intelligence close to the people, information and physical environments it serves, while showing how device makers can turn Acrab’s computing platform into complete agentic AI experiences.
For decades, personal computing advanced in predictable steps: faster processors, larger screens, more storage. Agent Box represents something else entirely—the first system designed not to run programs, but to host intelligence.
Agent Box brings together local language and vision model inference, multimodal interaction, persistent memory and an orchestration layer that can understand goals, break tasks into steps and coordinate action across agents, systems and connected devices. Users’ data and memories remain private and stored locally on the device, while the system grows more capable and customized as the context deepens and memories accumulate. Acrab designed the compute architecture from the ground up to achieve optimal local AI performance, usability, cost efficiency, and power efficiency within one device.
Acrab’s Agent Box
A purpose-designed SoC for large model inference at the edge
GΞLIX 1 is built on a 5-nanometer process and is Acrab’s first SoC designed specifically for edge AI. Rather than relying on separate compute components, it integrates CPU, GPU and NPU resources with a unified memory architecture engineered for large AI models and agentic workloads.
GΞLIX 1, Acrab’s 1st Generation edge AI system-on-chip (SoC)
The SoC features a 20-core Arm CPU, multicore NPU acceleration and 273 GB/s of unified memory bandwidth. It is designed to support local deployment of open-source models in up to the 100 billion parameter class, with coordinated execution across CPU, GPU and NPU resources. Supporting models at this scale locally places substantial demands on computing performance, memory bandwidth and power efficiency.
GΞLIX 1 is engineered for rapid responses at power levels suitable for systems that remain active throughout the day. A central design goal was reducing the delay before a model begins to respond, particularly with long prompts and large context windows.
In company testing, GΞLIX 1 achieved a prefill rate of 1416.8 tokens per second under a Gemma 26B A4B configuration with a 40K KV cache and a 10K token input, compared with 188.9 tokens per second on Mac Mini M4 Pro, representing up to 7.5X faster prefill performance. These capabilities turn a single chip into a versatile supercomputing platform for a wide range of applications.
A full-stack platform, from silicon to applications
Beyond the SoC, Acrab has built the software and system layers needed to turn local model inference into working agentic products. These include an optimized runtime and developer toolchain, agent operating system capabilities, reference designs and applications that help devices understand context, retain memory and coordinate real-world action.
Agent Box is the first expression of Acrab’s broader ambition to provide a horizontal computing foundation for agentic AI across a wide range of edge devices and intelligent systems.
Processing a substantial share of AI workloads locally can reduce dependence on metered cloud inference, lower recurring processing and data transfer costs, and avoid the delay involved in sending every interaction to a remote service. Cloud resources can still be used when a task requires them, allowing developers to choose the right balance between local and cloud execution.
Building a broader edge AI device ecosystem
Acrab plans to work with device manufacturers and developers to bring its computing platform into products including AI NAS systems, AI PCs, smart vehicles, and industrial and service robots.
Agent Box demonstrates how Acrab’s silicon and software can be integrated into a complete product experience. The company aims to provide a complete set of compute platform and agent-native infrastructure for the next generation of AI transformation across industries. By combining custom AI silicon, full-stack software, and reference designs of agents for use scenarios, Acrab enables industry partners and developers to bring intelligent AI products to market faster.
“Our goal is to give device makers and developers the foundation to bring agentic intelligence into many different products and environments,” Dr. Phua said. “Agent Box demonstrates what the technology can do today, while GΞLIX 1 and our full-stack platform are designed to support a much broader ecosystem of devices and applications.”
Product Launch Event Video Replay:
About Acrab
Acrab is a technology company building agentic AI compute infrastructure for the next generation of intelligent systems. Founded in 2024, the company develops high-performance AI compute architecture and integrated software platforms designed to bring AI agents into action, providing personalized assistance and real-time execution across a range of edge environments.
By combining purpose-designed silicon, advanced edge AI models, full-stack software and system orchestration, Acrab provides the computing foundation for AI agent systems across everyday life, bringing assistance, creativity, utility and value.
In June 2026, Acrab announced that it had received over US$350 million in cumulative financing from global venture capital firms and strategic industry investors, including early backers Vertex Ventures Southeast Asia & India, Vertex Growth, and K3.
For more information about Acrab, please visit https://www.acrab.ai/.
