NVIDIA has published a technical blog detailing the optimization of the Llama 3.2 collection of open models using NVIDIA NIM microservices to enhance flexible AI experiences. Additionally, Neural Magic has achieved a significant breakthrough by fine-tuning Llama-7B at 70% sparsity with zero accuracy loss and an 8.6x inference speedup. The blog also highlights the integration of NVIDIA's RTX-powered systems with Brave's AI assistant, Leo, which can summarize webpages, generate content, translate, and analyze text. Users can connect local AI models to the Brave browser using tools like Ollama, allowing them to run models locally and interact through a command-line window. More than 50 tools and apps are now accelerated with llama.cpp on the RTX platform, and the privacy-preserving Leo can also write code and create new text.