Today NVIDIA announced a new NVIDIA NIM™ Agent Blueprint. Each blueprint includes NVIDIA NIM and partner microservices, one or more AI agents, open-source sample code, customization instructions, and a Helm chart for deployment. Developers can modify these blueprints using proprietary data and deploy them anywhere—in the data centers and the cloud. Explore these blueprints:
Vulnerability Analysis for Container Security: Accelerate detection and mitigation of software vulnerabilities with AI agents powered by retrieval-augmented generation with NVIDIA NIM and Morpheus.
We’re excited to let you know that NVIDIA NIM™ inference microservices for the following models are available for self-hosted deployment on your choice of NVIDIA-accelerated infrastructure:
Maxine Eye Contact: Redirects the user’s eye gaze in real time, for video conferencing or post-production, to simulate direct eye contact with the camera and audience.
Mistral NeMo 12B: Multilingual language model for reasoning and code from Mistral AI and NVIDIA.
NV-CLIP: Generating vector embeddings for the given image or text.
RFdiffusion: Given specific design constraints, RFdiffusion can create completely novel or highly customized protein structures.
Additionally, the following models are now available to explore through a browser or with free credits for calling NVIDIA-hosted API endpoints in the NVIDIA API catalog—all powered by NIM:
AI Generated Image Detection: Robust image classification model for detecting and managing AI-generated content.
Edify-3D: Multimodal architecture, trained by Shutterstock on their licensed content, that generates ready-to-edit 3D meshes in two minutes.
Edify-360-HDRi: Multimodal architecture, trained by Shutterstock on their licensed content, that generates 360 HDRi for background and lighting of 3D scenes.
Llama 3.2 Vision Language Models: Cutting-edge VLMs (11B and 90B variants) excelling in high-quality reasoning from images.
Llama 3.2 Small Language Model: Advanced SML with language understanding, superior reasoning, and text generation.
Maxine Studio Voice: Maxine Studio Voice transforms the input speech recorded on low-quality microphones or in noisy/reverberant environments into studio-recorded quality speech.
Nemotron 70B Reward Model: Leaderboard topping reward model supporting RLHF for better alignment with human preferences.
NVIDIA Llama-3.1-Nemotron-51B: Unique language model that delivers an unmatched accuracy-efficiency performance.
Qwen 2 7B: Chinese and English LLM targeting language, coding, mathematics, and reasoning tasks.
Take advantage of your free NVIDIA credits and try out these models today on ai.nvidia.com. If you’re running out of free credits, you can request additional credits one more time. You can also get free access to downloadable NIM microservices for research, development, and testing through the NVIDIA Developer Program.
For more information, see A Simple Guide to Deploying Generative AI With NVIDIA NIM. Join developers in building a retrieval-augmented generation (RAG) application using NIM microservices and LlamaIndex, and compete for exciting prizes by registering for our NVIDIA and LlamaIndex developer contest.
Top comments (0)