NVIDIA just quietly gave every builder free access to 80 AI models. No one is talking about this. MiniMax M2.7. DeepSeek 3.2. Kimi 2.5. GLM 5.1. GPT-OSS-120B. All of them. Free inference. Hosted API. Right now. It plugs straight into Cursor, Claude Code, Zed IDE, and Hermes agent with 3 lines of setup. The people who find out about this today will be running experiments that cost everyone else $300 a month. For free. Grab your API key at http://build.nvidia.com/models Set your base URL to http://integrate.api.nvidia.com/v1 Pick a model and start building. The gap between people who know this and people who don't just got wider. Your call.(CyrilXBT)