Skip to content
Back to models
Model · NVIDIA
NVIDIA

Nemotron 3.5 Lightning

nvidia/nemotron-3.5-lightning
Try in Agents

NVIDIA's 30B hybrid Mamba-Transformer MoE (3B active) — a very fast, very cheap reasoning model for agents and tool calls

reasoningcaching
Specs
Context
262K
Max output
131K
Provider
NVIDIA
Input
$0.060/ 1M tokens
Output
$0.160/ 1M tokens
Cache read
$0.030/ 1M tokens
Providers and pricing
Vercel AI GatewayList priceOur price
Input$0.060 / 1M tokens$0.060 / 1M tokens
Output$0.160 / 1M tokens$0.160 / 1M tokens
Cache read$0.030 / 1M tokens$0.030 / 1M tokens

List price is what the provider charges us. Our price is what a request is billed.

Live performance
Loading measurements…
How we measure. Every request is timed from the moment we send it. A finish time comes from the provider's own timestamps when it reports them, from the moment it calls us back when it doesn't, and otherwise from polling: we take the midpoint between the last check that found the job running and the first that found it done, so it is off by at most half the poll interval. First token is the wait for the first streamed text or tool token; output speed counts tokens after that first token. Requests you stop are left out. Figures appear once a window has 20 completed requests, and are aggregate only.
Try it
Chat with Nemotron 3.5 LightningTerramind API
⌘+Enter to run
Code
import { streamText } from "ai";
import { createOpenAICompatible } from "@ai-sdk/openai-compatible";

const terramind = createOpenAICompatible({
  name: "terramind",
  baseURL: "https://terramind.com/api/v1",
  apiKey: process.env.TERRAMIND_API_KEY!,
});

const result = streamText({
  model: terramind("nemotron-3.5-lightning"),
  prompt: "Why is the sky blue?",
});

for await (const chunk of result.textStream) {
  process.stdout.write(chunk);
}

Want to run this in your app? Grab an API key · Full chat reference

More by NVIDIA
ModelContextInOutCapabilities
NVIDIAnvidia/nemotron-3-super256K$0.150$0.650
reasoning
NVIDIAnvidia/nemotron-3-ultra1M$0.600$2.40
reasoningcaching
NVIDIAnvidia/nemotron-3.5-super-vl1.0M$0.200$0.500
reasoningvisionvideo