NVIDIA NIM Free Inference Endpoints

Free NVIDIA NIM inference endpoints to test leading AI models

The catch:

Free endpoints are limited to catalog models tagged "Free Endpoint" and can be deprecated; you need a verified NVIDIA account and an API key, and no rate limits or SLA are published for the free tier.

Source announcement

NVIDIA: 1M-Context Nemotron 3.5 Super VL Free on OpenRouter

Overview

NVIDIA build offers free inference endpoints for developers to preview, prototype, and build applications with leading generative AI models. The platform allows developers to test community and proprietary models through standardized cloud APIs before deploying them on local workstations, DGX systems, or cloud infrastructure.

What is Included

Free Inference Endpoints: Run test inferences directly in the cloud on top-tier foundational and specialized architectures without provisioning your own compute.
Featured Free Models:
nvidia/nemotron-3.5-super-vl-120b-a12b: Vision-language reasoning model (text, image and video input) for agentic and coding work.
deepseek-ai/deepseek-v4.1-flash: Advanced MoE model optimized for complex reasoning.
nvidia/nemotron-3.5-lightning-30b-a3b: Customizable, fast enterprise-grade model.
nvidia/nemotron-3-ultra-550b-a55b: High-capacity model tailored for multi-step agentic workflows.
Agentic Skills & Blueprints: Access to pre-built capabilities across AI/ML, accelerated computing, developer tools, and full end-to-end application blueprints.
NemoClaw Support: Explore safe agent execution, access controls, and data protection playbooks directly in the workspace.

How to Get Started

Navigate to the NVIDIA build portal.
Log in or create an NVIDIA Developer account.
Select an inference model from the catalog to generate an API key and begin querying endpoints or experimenting in the interactive playground.

Models served free

Limits as the provider publishes them; a dash means none is published.

Model	Context	RPM	RPD	TPM	Tools	Vision
nemotron-3.5-super-vl-120b-a12bFree inference endpoint; text, image and video input	262k	—	—	—	yes	yes
deepseek-v4.1-flashFree inference endpoint; limits not published	1,048k	40	10,000	—	yes	yes
nemotron-3.5-lightning-30b-a3bFree inference endpoint; text input only	1,048k	40	10,000	—	yes	no
nemotron-3-ultra-550b-a55bFree inference endpoint	1,048k	40	10,000	—	yes	no

History

Verified by a person on Oct 12, 2026 · last automated check Oct 24, 2026: the page could not be read automatically. Automated findings never change this page until a person confirms them (how we verify).

Oct 21, 2026Terms updatedby a person · Content updated from https://build.nvidia.com/ (checked 2026-10-20).
Oct 12, 2026Verifiedby a person
Sep 10, 2026Listedby a person

Full history with screenshots · Sitewide changelog

Still works?

Report

Work at build.nvidia.com? Add an aifree.dev badge

NVIDIA NIM Free Inference Endpoints on aifree.dev

HTML

<a href="https://aifree.dev/offer/nvidia-nim-free-inference"><img src="https://aifree.dev/badge/nvidia-nim-free-inference" alt="NVIDIA NIM Free Inference Endpoints on aifree.dev" height="20"></a>

Markdown

[![NVIDIA NIM Free Inference Endpoints on aifree.dev](https://aifree.dev/badge/nvidia-nim-free-inference)](https://aifree.dev/offer/nvidia-nim-free-inference)