Nemotron 3 Ultra: A Daily Driver for Your Stack?

Оригинальный автор：Ray Fernando

Первоисточник：https://www.youtube.com/embed/v9snL5fTtLk

NVIDIA dropped Nemotron 3 Ultra today. It's the strongest open-weight model built in the US right now (550B total, 55B active), and it's fast. But the benchmark isn't the test I care about. The real question is whether an open model is good enough to be the daily driver in your agent stack. So we're setting it up live in OpenClaw and Hermes Agent and running real tasks. No demos. First job is finding the inference provider that actually runs it best (build.nvidia.com, OpenRouter, DeepInfra, Fireworks, Nebius Token Factory, whoever wins). Then we throw real agentic work at it and watch where it holds up and where it breaks. I haven't set up Hermes yet, so we might run the OpenClaw to Hermes migration live too. Could be fun. Come hang out in my free community, that's where I'm dropping all the resources and updates: https://rfer.me/startmyai GET THE TOOLS Nemotron 3 Ultra weights: https://huggingface.co/nvidia NVIDIA's Ultra breakdown for long-running agents: https://developer.nvidia.com/blog/nvidia-nemotron-3-ultra-powers-faster-more-efficient-reasoning-for-long-running-agents/ Run it via build.nvidia.com or OpenRouter: https://openrouter.ai Hermes Agent (Nous Research): https://github.com/NousResearch/hermes-agent OpenClaw to Hermes migration guide: https://hermes-agent.nousresearch.com/docs/guides/migrate-from-openclaw OpenClaw: https://openclaw.ai/ What we're testing live Setting up Nemotron 3 Ultra and finding the fastest, cheapest inference provider Running real agentic tasks in OpenClaw and Hermes Agent Whether it's good enough to be a daily driver, not just a benchmark winner Possible live OpenClaw to Hermes migration Timestamps by https://raytranscribes.com 00:00:00 Nemotron 3 Ultra for Agentic Coding 00:01:16 Nemotron 3 Ultra Providers 00:03:10 Live Research Task with Cursor 00:04:05 Fireworks AI Hosting Details 00:04:47 Open Weights vs. Open Source 00:07:30 Fireworks AI Model Configurations 00:11:06 Free Access via Nous Research 00:11:58 Live Hermes Agent Installation 00:15:18 Hermes Agent Setup Bugs 00:20:46 Hermes Agent UI First Look 00:22:25 Subscribing to Free Nous Portal 00:24:12 Configuring the Free Model 00:26:36 First Image Analysis Test 00:29:00 Nutrition Label Image Test 00:30:26 Bug: Hermes Agent Image Handling Failure 00:33:09 Auxiliary Vision Model Investigation 00:38:20 Cursor's Provider Research Results 00:41:15 Nebius Token Factory Speed Test 00:46:06 Recipe Generation Test on Nebius 00:52:00 OpenClaw Configuration Draft 00:54:24 Personal Agentic Workflow with OpenClaw 00:56:29 Live Demo of "Pebbles" Voice Agent 00:58:16 Live Multilingual Capabilities Test 01:02:00 The "Pebbles" App Design Concept 01:05:00 Philosophy of Building AI Software 01:08:43 The Origin Story of "Pebbles" 01:12:47 Evolving the App Design 01:16:41 The Importance of User Experience 01:18:02 Anker Prime Power Station 01:19:00 Final Thoughts CONNECT WITH RAY X (Twitter): https://x.com/RayFernando1337 #Nemotron3 #OpenClaw #HermesAgent #AICoding #OpenWeights