Nemotron 3.5 Lightning now available on AI Gateway for free
NVIDIA’s Nemotron 3.5 Lightning reasoning model is now freely accessible via Vercel’s AI Gateway, offering a 30B-parameter hybrid architecture with a 1M token context window and tool-calling support.
Vercel has made NVIDIA’s Nemotron 3.5 Lightning available at no cost on its AI Gateway platform under the identifier nvidia/nemotron-3.5-lightning-free, with inference served directly by NVIDIA. The model is positioned as an open-weight reasoning and chat model optimized for long-running agents, sub-agent workflows, chatbots, and retrieval-augmented generation systems. Its hybrid Mixture-of-Experts design combines 30 billion total parameters with 3 billion active per token, integrating Mamba-2 and MoE layers alongside a limited number of attention layers.
Nemotron 3.5 Lightning includes a 1 million token context window and supports tool calling and structured outputs, enabling integration with external functions and formatted data generation. The provider, NVIDIA, may retain prompts submitted to the model for purposes of model improvement, as noted in the announcement. Users can access the model through multiple interfaces, including the AI SDK, Chat Completions, Responses, and Messages APIs, or via coding agents connected to AI Gateway.
To deploy Nemotron 3.5 Lightning in supported coding environments such as Claude Code, Codex, and Cursor, users must install the latest Vercel CLI and complete a setup process. The setup involves selecting the model nvidia/nemotron-3.5-lightning-free within the agent interface. Vercel provides a dedicated guide for configuring coding agents to interact with models on AI Gateway, ensuring compatibility and performance optimization.
AI Gateway functions as a unified interface for model inference, usage tracking, and cost management, offering features such as configurable retries, failover mechanisms, and performance optimizations to enhance uptime beyond provider guarantees. The platform includes built-in custom reporting tools, API key budget controls, and routing rules. Pricing is passed through directly from providers with no additional markup, and no platform fee is charged on inference, including for Bring Your Own Key requests. A full catalog of available language models is accessible via the AI Gateway interface.