Free APILLMAPI keyOfficial
NVIDIA NIM (build.nvidia.com)
100+ optimized models free for prototyping at build.nvidia.com
NVIDIA's hosted NIM catalogue at build.nvidia.com exposes 100+ models (Nemotron, Llama, DeepSeek, Qwen, GLM, plus vision and retrieval models) behind an OpenAI-compatible endpoint. Signing up for the free NVIDIA Developer Program gives you an nvapi- key with no credit card for prototyping.
Free tier
Free for prototyping via the NVIDIA Developer Program (no card). Governed by a per-minute rate limit rather than a per-token price; production hosting is separate.Rate limits
Free hosted access is ~40 requests/minute (verified June 2026), with an application path toward ~200 RPM. The old signup credit balance was phased out (early 2025); treat the 40 RPM rate limit as the real ceiling.Auth
API keyBase URL
https://integrate.api.nvidia.com/v1What you'd build with it
- Try Nemotron and large open models (DeepSeek-R1, GLM) without GPU setup
- Embeddings + reranking for RAG using NeMo Retriever models
- Vision-language tasks against NVIDIA-optimized VLMs
- A free prototyping backend before self-hosting the same NIM containers
Key endpoints
- POST
/chat/completionsOpenAI-compatible chat completion across catalogue models - POST
/embeddingsGenerate embeddings from NeMo Retriever models - GET
/modelsList available hosted NIM models
Members
Members get the runnable recipe
Sign in free to copy the example request, see a sample response, and read the gotchas.
Tags
llminferenceopenai-compatiblevisionrag