Vercel adds NVIDIA's Nemotron 3.5 Lightning to AI Gateway for free
Vercel has made NVIDIA's Nemotron 3.5 Lightning available for free on its AI Gateway under the model name nvidia/nemotron-3.5-lightning-free. According to Vercel, the open-weight reasoning and chat model is served by NVIDIA and aimed at long-running agents, sub-agent work, chatbots and RAG systems.
- Vercel lists NVIDIA's Nemotron 3.5 Lightning for free on AI Gateway as nvidia/nemotron-3.5-lightning-free.
- The open-weight model has 30B parameters with 3B active per token and a hybrid Mamba-2 and Mixture-of-Experts design.
- It offers a 1M token context window, tool calling and structured outputs.
- The provider may retain prompts for model improvement.
- AI Gateway reflects provider pricing with no markup and no platform fee on inference, including BYOK.
What Nemotron 3.5 Lightning offers
Nemotron 3.5 Lightning is described by Vercel as an open-weight model built for reasoning and chat. The listed use cases are long-running agents, sub-agent work, chatbots, and retrieval-augmented generation (RAG) systems.
On the technical side, the model has 30B parameters, with 3B active per token. It uses a hybrid Mixture-of-Experts design that interleaves Mamba-2 and MoE layers with a few attention layers. It offers a 1M token context window and supports tool calling and structured outputs.
One caveat applies to the free offering: according to Vercel, the provider may retain prompts for model improvement. Teams handling sensitive data should take that into account before sending requests.
How to call the model
Developers can reach Nemotron 3.5 Lightning in several ways. Vercel lists the AI SDK, the Chat Completions, Responses and Messages APIs, and coding agents connected to AI Gateway. In every case, the model name to use is nvidia/nemotron-3.5-lightning-free.
Vercel's example uses the AI SDK's streamText function with that model name and a prompt asking for a TypeScript function that retries a failed fetch with exponential backoff.
For coding agents such as Claude Code, Codex and Cursor, Vercel says to install the latest Vercel CLI and run npx vercel ai-gateway setup. After that, the model can be selected inside the agent. Vercel points to its coding agents guide for details.
What AI Gateway provides
AI Gateway is Vercel's unified API for calling models. According to the company, it also tracks usage and cost and lets teams configure retries, failover and performance optimizations for higher-than-provider uptime. Built-in features include custom reporting, budgets for API keys and routing rules.
On pricing, Vercel states that AI Gateway reflects provider pricing with no markup and does not charge a platform fee on inference. That also applies to Bring Your Own Key (BYOK) requests. The full list of language models on the platform can be viewed on the AI Gateway site.
What it means for developers
For developers already using AI Gateway, the listing adds a no-cost option with a very large context window and support for tool calling and structured outputs. That combination suits experiments with agents and RAG pipelines without a separate provider setup, as the model sits behind the same API as other models on the platform.
The trade-off is the data retention note. Since the provider may keep prompts to improve the model, the free tier fits testing and non-sensitive workloads better than confidential ones, based on what Vercel has disclosed.
