Intel: 500 AI Models Have Been Optimized For Core Ultra Processors

The semiconductor giant says its AI model optimization work is important because ‘models form the backbone of AI-enhanced software features like object removal, image super resolution or text summarization.’

Intel said more than 500 AI models have been optimized for its Core Ultra processors, calling the move a “significant milestone” in its effort to position itself as the prime chip supplier for AI PCs.

The Santa Clara, Calif.-based company said on Wednesday that the AI models span “more than 20 categories of AI, including large language, diffusion, super resolution, object detection and computer vision.” These models are accessible through industry sources such as Hugging Face, ONNX Model Zoo, OpenVINO Model Zoo and PyTorch.

[Related: AMD Says Ryzen Pro CPUs Enable Better AI PCs For Businesses]

They include Microsoft’s Phi-2 small language model, Meta’s Llama large language model, OpenAI’s Whisper speech recognition model, Stability AI’s Stable Diffusion 1.5 text-to-image generation model, Google’s Bert natural language understanding model and the Mistral language model.

Intel said its optimization work is important because “models form the backbone of AI-enhanced software features like object removal, image super resolution or text summarization.” The models can be used across the Core Ultra’s CPU, GPU and neural processing unit (NPU), it added.

“There is a direct link between the number of enabled/optimized models and the breadth of user-facing AI features that can be brought to market,” the company said. “Without a model, the feature cannot be designed. Without runtime optimization, the feature cannot reach its best performance.”

The semiconductor giant is in an arms race with rivals AMD and Qualcomm to not only provide the best processors for AI PCs but also to enable compelling software experiences with the goal of creating greater demand for their respective products.

Outside of the AI model optimization project, Intel has been working with more than 100 independent software vendors (ISVs) on more than 300 AI-powered features for PCs powered by Core Ultra processors. This is being done as part of the company’s AI PC Acceleration Program that was established a few months before it released the Core Ultra lineup in December.

The company said its software enablement work is made possible by the investments it has made into client AI processing, framework optimizations, AI tools like OpenVINO and the broader work that has been required to establish AI PCs as a new device category.

“This unmatched selection reflects our commitment to building not only the PC industry’s most robust toolchain for AI developers, but a rock-solid foundation AI software users can implicitly trust,” said Robert Hallock, vice president and general manager of AI and technical marketing in Intel’s Client Computing Group, in a statement.