Nemotron 3.5 Lightning now available on AI Gateway for free
Source
vercel.com
Author
Zachary Chen
Date
Why it matters
A free 1M-context open-weight MoE model with tool calling is now callable through AI SDK and coding agents via AI Gateway. Note the provider may retain prompts.
Key takeaways · AI-distilled
Vercel's AI Gateway now serves NVIDIA's Nemotron 3.5 Lightning at no cost under the model ID nvidia/nemotron-3.5-lightning-free, with NVIDIA itself acting as the serving provider.
Vercel describes the model as open-weight with 30B parameters and 3B active per token, using a hybrid design that interleaves Mamba-2 and mixture-of-expertsA model built from many specialist sub-networks where only a few activate per token, giving big-model capability at small-model running cost.Full definition → layers with a few attentionThe mechanism that lets a model weigh which earlier words matter for the word it's currently processing — the core operation of a transformer.Full definition → layers, plus a 1M token context windowThe maximum amount of text a model can consider at once — its working memory for the current conversation or task.Full definition →, tool useA model's ability to call external functions — run code, search the web, edit files — instead of only generating text.Full definition → and structured outputs.
The free endpoint carries a data caveat: Vercel's changelog notes that the provider may retain prompts for model improvement.
It is callable through the AI SDK and the Chat Completions, Responses and Messages APIs, and running npx vercel ai-gateway setup connects coding agents such as Claude Code, Codex and Cursor to it.
Terms in this piece · Glossary
mixture-of-experts — A model built from many specialist sub-networks where only a few activate per token, giving big-model capability at small-model running cost.
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
tool use — A model's ability to call external functions — run code, search the web, edit files — instead of only generating text.
attention — The mechanism that lets a model weigh which earlier words matter for the word it's currently processing — the core operation of a transformer.
