NVIDIA x Red Hat Small Models Hack

Small. Sharp. Specialized. Build AI that punches above its weight—specialized small models, multi-agent systems, and reliable tool-calling—all running on NVIDIA Brev or locally. Join NVIDIA and Red Hat in Raleigh for the Small Models Hack. This event brings together frontier builders, applied ML engineers, and systems thinkers to explore what's possible when small, open models (total parameters must be ≤ 33 billion) take center stage. Every team will get NVIDIA Brev credits to spin up GPU environments throughout the weekend. ​✏️ How to Participate This hackathon is limited to a maximum of 100 participants. Once we reach capacity, additional applicants will be placed on a waitlist and notified if a spot becomes available. Application Guidelines: ​🔥 Challenge Tracks Teams can tackle any real-world problem they care about. Each track defines how your AI works, not what problem it solves. 🎯 1. Fine-Tuned & Specialized Focus: A small model that's excellent at one real thing. The Goal: Adapt or fine-tune a small model for a specific job—then publish it openly so others can reuse it. Depth over breadth, and evidence that specialization beats general-purpose for your chosen task. 🕸️ 2. Multi-Agent Systems Focus: Many small models beating one big one. The Goal: Wire several small models together so they coordinate—dividing up work, passing context back and forth, building on each other's output—to accomplish something no single model could alone. 🛠️ 3. Tool-Calling Focus: A single model that uses real tools reliably. The Goal: Deciding when to call, picking the right tool, formatting valid arguments, handling what comes back, recovering from failed or malformed calls, verifying its own work, and falling back gracefully when something breaks. ​💰 What You Can Win Qualifying submissions, depending on the unique track requirements, are eligible to win the following and more: NVIDIA Privacy Policy: https://www.nvidia.com/en-us/about-nvidia/privacy-policy/