100+ AI Models, Free

The Idea

The best open AI models in the world — DeepSeek, Kimi, GLM, MiniMax, Qwen, even OpenAI's open-weight models — are now hosted by NVIDIA and free to use through one API key. Open-weight always meant "free if you can run it," and running a frontier model takes serious GPUs. NVIDIA just took that part off your plate.

It's OpenAI-compatible, so it drops into whatever you already build with by changing two lines — the base URL and the key. No credit card, no trial timer.

Set It Up

1. 1.Go to build.nvidia.com and sign up — email or an NVIDIA account, no card.
2. 2.Open Settings → API Keys, generate a key (it starts with nvapi-), and copy it — it's shown once.
3. 3.Paste the prompt below into Claude and let it wire the rest into your project.

Doing it by hand? It's the OpenAI SDK with two values swapped — base URL https://integrate.api.nvidia.com/v1 and your nvapi- key. The same two values plug into any tool with a custom OpenAI base URL — Cursor, Cline and most OpenAI-compatible setups.

The Headliners

Over 100 models in the catalog. The ones builders actually reach for:

New to it? Start with DeepSeek — strong all-rounder. Then rotate through the rest; having 100+ behind one key is the whole point. Copy each model's exact ID from its page on build.nvidia.com — they look like deepseek-ai/deepseek-v3.1 and do change over time.

The Honest Bit

The free tier is for building, testing and prototyping — not running a high-traffic app. Expect a starter pool of credits and a per-minute rate limit, free during the preview period. That's plenty to build and test almost anything. Just don't architect a scaled product assuming infinite free calls — if real traffic shows up, you move to a paid tier or self-host.