Intel continues to invest in its AI software ecosystem to support its “AI Everywhere” strategy, ensuring new models are optimized for Intel’s AI hardware.

Intel announced on the 25th that Meta's latest LLM Rama 3.1 provides performance data and optimizations across data center, edge, and client AI products.

Following the launch of the Rama 3 last April, Meta has released the Rama 3.1, its most powerful model to date. Rama 3.1 offers several new updated models in various sizes and features, including the Rama 3.1 405B (45 billion-piece model), the largest foundation model publicly available.

These new models are enabled and optimized on Intel AI products via open ecosystem software including PyTorch and Intel PyTorch Extensions, DeepSpeed, Hugging Face Optimum Library, and vLLM.

We also support OPEA (Open Platform for Enterprise AI), a new open platform project from LF AI and the Data Foundation to build open, multi-vendor, powerful, and composable generative AI solutions that leverage the best innovations from across the ecosystem.

The Llama 3.1 Multilingual LLM Collection is a collection of pre-trained and tuned generative models in sizes 8B, 70B, and 405B, all supporting long context lengths across eight spoken languages. Llama 3.1 405B has state-of-the-art capabilities in general knowledge, operability, mathematics, tool use, and multilingual translation.

Intel's AI platform and solutions accelerate the deployment of AI RAG for enterprises. As one of the founding members of OPEA, Intel is leading the way in building an open ecosystem for AI for enterprises. It is noteworthy that OPEA has optimized performance by utilizing the Rama3.1 model.

Currently, Intel AI PC and data center AI product portfolio and solutions can run Rama 3.1, and OPEA is fully enabled with Rama 3.1 on Intel Gaudi 2 and Xeon families. Intel is continuously optimizing the software to support new models and use cases.