The GenAI juggernaut continues, as Meta and partners celebrated Meta AI powered by Llama 3 on its integration into multiple platforms and released on more global markets.
Meta AI is a rival to OpenAI’s ChatGPT and can now be deployed across Facebook, Instagram, WhatsApp and Messenger, as well as a standalone website, Meta AI. The company’s stock rose 3% Thursday on the integration news.
Meta AI in English is being released in a dozen countries outside the US including Australia and Canada.
The integration of Llama 3 provoked a round of announcements from big hardware names including AMD, AWS, Dell, Intel, Nvidia and Qualcomm. Llama 3 models are also expected soon ono AWS, Databricks, Google Cloud, Hugging Face, Kaggle, IBM Watson X, Microsoft Azure, Nvidia NIM and Snowflake. Nvidia NIM, announced in March, includes a generative text-to-image model designed to synthesize photorealistic images from a text prompt in a single network
Nvidia said it trained Llama 3 on a computer cluster made up of 24,576 Nvidia H100 Tensor Core GPUs linked with its Quantum-2 InfiniBand network. Meta then tuned its network, software and model architectures for Llama 3. Meta also has described plans to scale its infrastructure to 350,000 H100 GPUs. Developers can try Llama 3 at ai.nvidia.com, where it’s packaged as an Nvidia NIM microservice.
Llama 3 runs with Nvidia AI Enterprise support and also runs on Nvidia Jetson Orin for robotics and edge devices where commands can be used for adjustments to their operations, for example, along the lines CEO Jensen Huang recently described at GTC2024
RELATED: Huang posits Nvidia has moved well beyond GPUs, but isn’t a CSP
Qualcomm and Intel also jumped on the Llama 3 bandwagon. Qualcomm said it will collaborate with Meta to optimize Llama 3 on smartphones, PCs, VR/AR headsets, vehicles and more. Llama 3 will provide superior responsiveness, enhanced privacy, better reliability and more personalized experiences, Qualcomm said.
For its part, Intel said its Gaudi, Xeon, Core Ultra processors and Arc graphics for data center, edge and AI PCs will offer developers the latest optimizations to run Llama 3. Gaudi 2 accelerators have optimized performance for Llama 2 and now have initial performance measurements for Llama 3, Intel said.
Meta launched its previous version, Llama 2, in July 2023. News of Llama 3 integrations pushed Meta’s stock up 3% on Thursday but Nvidia only got a bump less than 1% and Intel saw a decline of 2% while Qualcomm dropped 1.6%.