Vercel's AI Gateway adds a free, 1M-token open-weight model

Vercel added Nvidia’s Nemotron 3.5 Lightning to AI Gateway on 14 October, free to call as nvidia/nemotron-3.5-lightning-free. It’s an open-weight reasoning and chat model aimed at long-running agents, sub-agent work, chatbots and RAG systems.

The model has 30 billion total parameters with 3 billion active per token, a hybrid Mixture-of-Experts design that interleaves Mamba-2 and MoE layers with a handful of attention layers. It carries a 1 million token context window and supports tool calling and structured outputs. You call it through the AI SDK, the Chat Completions, Responses or Messages APIs, or any coding agent wired up to AI Gateway, with one API key covering spend tracking and request traces:

import { streamText } from 'ai';
const result = streamText({
  model: 'nvidia/nemotron-3.5-lightning-free',
  prompt: 'Summarise this 400-page contract and flag the renewal clauses.',
});

Why it matters: a 1M-token context window with tool calling, at zero cost and no markup from Vercel, makes this a reasonable default for agent work that would otherwise burn through a paid context budget fast, especially sub-agent calls that don’t need a frontier model.

The caveat: Nvidia may retain prompts sent to the free tier for model improvement, so it’s not the one to route customer data or anything sensitive through.