Nemotron 3.5 Lightning free API on Vercel AI Gateway and OpenRouter at $0
NVIDIA's open Nemotron 3.5 Lightning has free versions on Vercel AI Gateway and OpenRouter, with $0 input and output. It has a 1M-token context, up to about 65K output tokens and tool calling. No end date has been announced, and the provider may keep prompts and responses from the free version for training.
The briefing
NVIDIA's new Nemotron 3.5 Lightning is now a free model on Vercel AI Gateway, and OpenRouter has a free version too. It costs nothing to call, which makes it a good fit for agent sub-tasks, chatbots and RAG.
What you can use
- On Vercel AI Gateway the model ID is
nvidia/nemotron-3.5-lightning-free, billed at $0 for input and output. - On OpenRouter the model ID is
nvidia/nemotron-3.5-lightning:free. It is also free, with rate limits on free endpoints. - The context window is 1M tokens with up to about 65K output tokens, and it supports tool calling. Vercel AI Gateway also supports structured outputs.
- No end date has been announced.
- The provider may retain prompts and responses from the free version to improve the model.
Specs and public benchmarks
Nemotron 3.5 Lightning is an open-weight mixture-of-experts model with 30B total parameters and 3B active per token. It interleaves Mamba-2 and MoE layers with a few attention layers for high throughput. On Artificial Analysis it scores 74.3% on GPQA Diamond, 60.3% on AA-LCR long-context reasoning and 24.3% on Terminal-Bench 2.1, so it suits long-document Q&A and light sub-tasks better than heavy coding work.
How to use
- Enable AI Gateway in Vercel, create an API key and set the model to
nvidia/nemotron-3.5-lightning-free. - Create an OpenRouter API key and set the model to
nvidia/nemotron-3.5-lightning:free. The API is OpenAI-compatible.
Details: Vercel AI Gateway model page · OpenRouter model page.