NVIDIA API Catalog lets you call 40+ models for free, no per-token billing

NVIDIA's API catalog at build.nvidia.com hosts dozens of models you can call for free. Sign up, generate an API key and start prototyping. There is no card required and no per-token billing.

What you can use

• More than 40 models currently have free endpoints, including DeepSeek V4.1 Flash, MiniMax M3.1 Flash, GLM-5.3-Flash, StepFun Step 5 Preview and NVIDIA's own Nemotron 3.5 text, vision, embedding and reranking models.
• The free tier is rate limited, up to 40 requests per minute for most models, with no per-token charges. Your personal limit is shown in the top-right corner of the dashboard.
• When you need more throughput, switch the same model to a partner endpoint or self-host the NIM container on your own GPUs. The API stays the same.

How to use it

1. Go to build.nvidia.com, create and verify your NVIDIA account, and join the free Developer Program.
2. Filter the model list for Free Endpoint, try a model in the browser, then generate an API key for your project.
3. Free endpoints are for prototyping, research and testing. Production use serving real users needs an NVIDIA AI Enterprise license or a partner endpoint.

See the NVIDIA NIM FAQ for the full terms.

Some details are based on StartupPerks (CC BY 4.0).