Innodisk Launches Edge AI Systems for Real-Time LLM and VLM Performance

Innodisk, a leading global provider of AI solutions, has unveiled a new generation of edge AI systems designed to bring real-time performance to large language models (LLMs) and vision language models (VLMs). These systems, powered by Intel Core Ultra Series 3 processors, are aimed at expanding the capabilities of AI workloads at the edge without relying on cloud infrastructure.

The new product line, including the APEX-E300, APEX-P300, and ASBC-3160, delivers up to 180 total platform TOPS (Tera Operations Per Second) of heterogeneous AI performance, making them suitable for demanding edge inference tasks. These systems are designed to address the growing need for advanced AI applications in various industries, including manufacturing, retail, and smart infrastructure.

According to Don Yu, Special Assistant to the General Manager at Innodisk, the new products offer organizations the flexibility to deploy sophisticated VLM and LLM applications directly at the edge. This capability is crucial for environments requiring immediate response, as cloud-dependent architectures can introduce latency, connectivity issues, and additional infrastructure costs.

"Organizations increasingly need to run multiple AI workloads simultaneously while maintaining tight control over power, performance, and latency. Our latest Intel Core Ultra Series 3 powered platforms give customers the flexibility to deploy complex AI applications directly at the edge and scale compute resources based on the needs of each use case."

The new products are built on a heterogeneous computing architecture that combines CPU, GPU, and NPU resources. This architecture enables organizations to run multiple AI models on a single edge platform while balancing performance, power consumption, and application requirements. For applications where power efficiency is critical, organizations can run AI workloads using the processor's integrated compute resources at power levels as low as approximately 25 watts, avoiding the additional power requirements associated with a discrete GPU.

For more demanding applications requiring greater model accuracy or larger AI models, the APEX-P300 provides additional scalability through PCIe expansion. This allows customers to tailor compute resources to specific workloads rather than deploying discrete GPU acceleration where it is not required.

Innodisk's new edge AI systems are designed to address emerging multi-model edge AI applications across various industries. In a quick-service retail environment, for example, a VLM could analyze visual information locally while an LLM simultaneously generates personalized recommendations or responses. In manufacturing, multiple models can be consolidated onto a single platform to support applications such as machine vision, inspection, anomaly detection, and intelligent automation.

Dan Rodriguez, corporate vice president and general manager of the Edge Computing Group at Intel, highlighted the importance of these new platforms. "Intel Core Ultra Series 3 is built exactly for this need, combining CPU, GPU, and NPU capabilities into a single chip to enable power-efficient AI. Innodisk's platforms extend those capabilities to customers deploying VLMs, LLMs, and agentic AI in real-world environments at the edge."

The new products are part of Innodisk's broader strategy to expand beyond the industrial market and pioneer innovations in edge AI domains. From storage to cutting-edge AI applications, Innodisk's solutions serve as reliable building blocks for AI deployments, enabling real-time data sensing, accelerated computing, and intelligent decision-making at the edge. Through hardware-software integration, Innodisk creates comprehensive solutions that address the dynamic needs of modern enterprises.