NVIDIA releases AI Foundation models for RTX AI PC and its CEO says AI is advancing at an "incredible pace"
NVIDIA CEO, Jensen Huang, introduced NVIDIA Cosmos, Blackwell RTX series 50 GPU, and artificial intelligence tools for PCs during his presentation at CES 2025.
Without a doubt, the conference of Jensen Huang, CEO of NVIDIA was one of the most anticipated at CES 2025, due to the great year the tech company had, where its shares registered a rise of 171%, making it the second most profitable company in the S&P 500 index, thanks to a strong demand for its chips used for artificial intelligence.
And those who arrived at the Las Vegas Nevada Convention Center did not leave empty-handed. In 90 minutes, the duration of his keynote at the kickoff of CES 2025, he made announcements that included new products for video games, autonomous vehicles, robotics, and AI agents.
"This is how he gave a summary of the rapid advancement of AI: "It all started with perceptual AI: understanding images, words, and sounds. Then, generative AI: creating text, images, and sounds"," said Huang. "Now we are entering the era of 'Physical AI,' an AI that can proceed, reason, plan, and act"."
Explaining that both NVIDIA's GPUs and platforms are at the center of these transformations, which are enabling advances in all industries, ranging from video games, robotics, and autonomous vehicles.
Follow us on: 
KNOW ALL THE DETAILS OF WHAT WAS PRESENTED AT CES
NVIDIA has just introduced basic models that will run locally on the computers with NVIDIA RTXu2122 AI which will empower the digital human, content creation, productivity, and development.
These models, which are offered as microservices NVIDIA NIMu2122, are processed with the new GPU GeForce RTXu2122 Series 50, which deliver up to 3.352 trillion operations per second of AI performance and 32 GB of VRAM. Based on NVIDIA's Blackwell architecture, the RTX 50 series GPUs are the first consumer GPUs to add support for FP4 computation, which doubles the AI inference performance and allows running generative AI models locally in a smaller memory space than the previous generation hardware.
GeForceu2122 has long been a vital platform for AI developers. The first GPU-accelerated deep learning network, AlexNet, was trained on the GeForce GTXu2122 580 in 2012, and last year, more than 30% of published AI research papers cited the use of GeForce RTX.
Now, with generative AI and RTX AI PCs, anyone can be a developer. A new wave of low-code and no-code tools, such as AnythingLLM, ComfyUI, Langflow, and LM Studio, allows enthusiasts to use AI models in complex workflows through simple graphical user interfaces.
The NIM microservices connected to these graphical interfaces will facilitate access and implementation of the latest generative AI models. The NVIDIA AI Blueprints built on NIM microservices, provide easy-to-use and preconfigured reference workflows for digital humans, content creation and much more.
To satisfy the growing demand of AI developers and enthusiasts, all major PC manufacturers and system assemblers are launching NIM-ready AI RTX PCs with GeForce RTX Series 50 GPUs.
"AI is advancing at the speed of light, from perception AI to generative AI and now to agentic AI," says Jensen Huang, founder and CEO of NVIDIA. "NIM microservices and AI Blueprints provide developers and PC enthusiasts with the building blocks to explore the magic of AI."
Make AI IncreNIMble
The basic models (neural networks trained from huge amounts of raw data) are the pillars of generative AI.
NVIDIA will launch a series of NIM microservices for the PC RTX AI from leading model developers such as Black Forest Labs, Meta, Mistral, and Stability AI. Use cases include large language models (LLM), vision language models, image generation, speech, embedding models for augmented retrieval generation (RAG), PDF extraction, and computer vision.
According to Robin Rombach, CEO of Black Forest Labs, "The GeForce RTX 50 series GPUs with FP4 calculation will unlock a vast array of models that can be run on the PC that were previously limited to large data centers. Converting FLUX into a NVIDIA NIM microservice increases the speed at which AI can be deployed and experienced by more users, while providing incredible performance."
NVIDIA has also just introduced the family of open models, called Nemotron, which provides high accuracy across a wide range of artificial intelligence tasks. The Llama Nemotron Nano model will be offered as a NIM microservice for PCs and RTX AI workstations, and it excels in agentive AI tasks such as instruction tracking, function calling, chatting, coding, and maths.
The NIM microservices include the key components to run AI on PCs and are optimized for deployment on NVIDIA GPUs, whether on RTX PCs and workstations or in the cloud.
Developers and enthusiasts will be able to download, configure, and quickly run these NIM microservices on Windows 11 PCs with Windows Subsystem for Linux (WSL).
"AI is driving innovation in Windows 11 PCs at an accelerated pace, and Windows Subsystem for Linux (WSL) provides an excellent multi-platform environment for AI development in Windows 11 along with Windows Copilot Runtime," says Pavan Davuluri, Corporate Vice President of Windows at Microsoft. "NVIDIA NIM microservices, optimized for Windows PCs, provide developers and enthusiasts with ready-to-integrate AI models for their Windows apps, further accelerating the deployment of AI capabilities for Windows users."
The NIM microservices, which run on the RTX AI PCs, will be compatible with the main development frameworks and AI agents, such as AI Toolkit for VSCode, AnythingLLM, ComfyUI, CrewAI, Flowise AI, LangChain, Langflow and LM Studio. Developers can connect applications and workflows created in these frameworks to AI models that run NIM microservices through standard industry endpoints, allowing them to use the latest technology with a unified interface in the cloud, data centers, workstations, and PCs.
Fans will also be able to experience a series of NIM microservices using an upcoming version of the tech demo NVIDIA ChatRTX.
Putting a Face to Artificial Intelligence
To demonstrate how enthusiasts and developers can use NIM to create AI agents and assistants, NVIDIA has just introduced Project R2X, a PC avatar with artificial vision capable of making information accessible to the user, assisting with desktop applications and video conferences, reading and summarizing documents, and much more.
The avatar is rendered using NVIDIA RTX Neural Faces, a new generative AI algorithm that enhances traditional rasterization with completely generated pixels. The face is then animated using a new Audio2Faceu2122-3D model based on diffusion that improves lip and tongue movement. R2X can connect to cloud AI services like OpenAI's GPT4 and xAI's Grok, and to NIM and AI Blueprints microservices, such as alternative PDF retrievers or LLMs, through development frameworks like CrewAI, Flowise AI, and Langflow.
The AI Blueprints Arrive at Your PC
The NIM microservices are also available for PC users through AI Blueprints, reference AI workflows that can run locally on RTX PCs. With these blueprints, developers can create podcasts from PDF documents, generate amazing images guided by 3D scenes, and much more.
The PDF to podcast blueprint extracts text, images, and tables from a PDF to create a podcast script that can be edited by users. It can also generate a complete audio recording from the script using voices available in the blueprint or based on a voice sample from the user. In addition, users can have a real-time conversation with the AI podcast host to get more information on specific topics.
The project uses NIM microservices such as Mistral-Nemo-12B-Instruct for language, NVIDIA Riva for text-to-speech conversion and automatic speech recognition, and the NeMo Retriever microservice collection for PDF extraction.
The AI Blueprint for 3D-guided generative AI gives artists more precise control over image generation. Although AI can generate stunning images from simple text instructions, controlling image composition using only words can be complicated. With this project, artists can use simple 3D objects arranged in a 3D renderer like Blender to guide the AI's image generation. The artist can create 3D assets by hand or generate them via AI, place them in the scene, and adjust the 3D view's camera. Then, a prepackaged workflow powered by the FLUX NIM microservice will use the current composition to generate high-quality images that match the 3D scene.
The NVIDIA NIM microservices and AI Blueprints will be available starting in February with initial hardware support for the professional GPUs GeForce RTX 50 Series, GeForce RTX 4090 and 4080, and NVIDIA RTX 6000 and 5000. In the future, more GPUs will be supported. NIM-ready RTX AI PCs will be available through Acer, ASUS, Dell, GIGABYTE, HP, Lenovo, MSI, Razer and Samsung, and from local system manufacturers Corsair, Falcon Northwest, LDLC, Maingear, Mifcon, Origin PC, PCS, and Scan.