Innodisk Brings Real-Time LLM and VLM Performance to the Edge with Intel Core Ultra Series 3 Processors

Innodisk, an AI solution provider, announced a new generation of Edge AI systems powered by Intel Core Ultra Series 3 processors, expanding its portfolio of high-performance, power-efficient platforms for AI workloads at the edge.

The new product line gives developers and enterprises greater flexibility to deploy Vision Language Models (VLMs), LLMs and other AI applications without relying on cloud infrastructure. At AI Infra Summit 2026, Innodisk demonstrated this flexibility through a live Agentic AI demo running on the APEX-E300 Edge AI System.

Cloud-dependent AI architectures can introduce latency, connectivity requirements and additional infrastructure costs that limit their effectiveness in environments requiring immediate response. Innodisk’s new products bring advanced AI inference directly to the edge, enabling applications to analyze data and respond locally in real time.

“Edge AI is rapidly evolving beyond running a single model for a single task. Organizations increasingly need to run multiple AI workloads simultaneously while maintaining tight control over power, performance and latency,” said Don Yu, special assistant to the general manager, Innodisk. “Our latest Intel Core Ultra Series 3 powered platforms give customers the flexibility to deploy sophisticated VLM and LLM applications directly at the edge and scale compute resources based on the needs of each use case.”

The new products are designed to address emerging multi-model Edge AI applications across industries including manufacturing, retail and smart infrastructure. 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.

“Organizations are increasingly running more advanced AI workloads at the edge,” said Dan Rodriguez, corporate VP and GM, edge computing group, Intel. “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.”

Flexible AI Performance from 25W Edge Computing to GPU-Accelerated Workloads
Built on a heterogeneous computing architecture that combines CPU, GPU and NPU resources, the Innodisk APEX-E300 and APEX-P300 leverage Intel Core Ultra Series 3 processors to intelligently distribute AI workloads across available compute engines. This architecture enables organizations to run multiple AI models on a single edge platform while balancing performance, power consumption and application requirements, delivering up to 180 total platform TOPS of AI performance.

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 25W, 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.