Try NIM

A small companion to this post. Paste your NVIDIA NIM API key, pick a model, type a prompt, and copy the command into your terminal. Your key never leaves your browser. This page does not make API calls; it generates a request you run locally, so the only host that ever sees your key is NVIDIA.

NIM API key

Stored only in this browser's localStorage on this domain.

Model nvidia / nemotron-3-super-120b-a12b, agentic flagship, 1M context openai / gpt-oss-120b, open-weights reasoning deepseek-ai / deepseek-v4-pro, coding, 1M context moonshotai / kimi-k2-thinking, reasoning minimaxai / minimax-m2.7 z-ai / glm5.1 qwen / qwen3-coder-480b-a35b-instruct, code flagship meta / llama-3.3-70b-instruct

Prompt

Run the command in any terminal. The first time, you may need pip install openai for the Python version. The curl version needs nothing beyond curl, which ships with macOS, Linux, and recent Windows.