At Extreme Investor Network, we are excited to share the latest developments in generative AI that are revolutionizing industries. NVIDIA, a leader in technology innovation, has recently introduced two groundbreaking model families, Phi-3 and Granite Code, as part of its NVIDIA AI Foundation.
Phi-3 Language Models
The Phi-3 series, in partnership with Microsoft, offers small language models (SLMs) that are optimized for superior performance and computational efficiency. These models excel in tasks like content generation, summarization, question-answering, and sentiment analysis, making them versatile for a range of applications requiring logical deductions and precise responses.
Phi-3 Vision Model
Additionally, the Phi-3 family includes the Phi-3 Vision model, a multimodal model with 4.2 billion parameters designed to process both text and visual data. This model, supporting 128K tokens, can analyze intricate visual elements in images, such as charts, graphs, and tables, making it ideal for data-rich tasks.
Granite Code
IBM has contributed the Granite Code models, open programming models tailored to assist with various coding tasks. Trained on 116 programming languages, these models can generate code examples, detect and correct errors, and offer explanations for code segments. Their performance on coding benchmarks is top-notch, and they are trained on license-permissible data, making them ideal for enterprise usage.
Optimized for Performance
Both Phi-3 and Granite Code models are optimized for latency and throughput using NVIDIA TensorRT-LLM. These models, alongside over three dozen other popular AI models supported by NVIDIA NIM, a microservice simplifying the deployment of performance-optimized models at scale, are driving significant advancements in AI transformations for enterprise application developers.
NVIDIA’s collaboration with leading model builders ensures that these models are fully supported on an accelerated stack, guaranteeing optimal performance and seamless deployment.
Get Started
For developers eager to experience, customize, and deploy these models in enterprise applications, we invite you to explore our API catalog. With complimentary NVIDIA cloud credits, you can begin testing these models at scale, connecting your applications to the NVIDIA-hosted API endpoint operating on a fully accelerated stack.
Image source: Shutterstock