NVIDIA Nemotron 3.5 Lightning is an open mixture-of-experts (MoE) model with 3 billion active parameters out of 30 billion total, built for high-throughput agentic workloads and specialized task execution. It supports a 1,000,000-token context window and returns text output, making it well suited for long-running agent loops, tool-heavy pipelines, and workflows that need to retain large amounts of context without the compute cost of a fully dense model.
By activating only 3B of its 30B total parameters per token, Nemotron 3.5 Lightning delivers throughput closer to a small dense model while drawing on a much larger parameter pool — useful for high-volume, latency-sensitive agentic deployments where cost per request matters.
The model supports file input, parallel tool/function calling, streaming responses, and structured outputs, making it a solid fit for agent frameworks and automated pipelines that call external tools and expect predictable, machine-parseable responses.
This model is currently offered free of charge — both input and output tokens are billed at $0 per 1M tokens.
Released August 11, 2026.
Measured on the AI/ML API gateway across live traffic. Figures are medians unless stated. Last updated 2026-10-12.
Nemotron 3.5 Lightning is documented as suited to code review, helping specialized agent workflows in software development. It is also documented for tool use, especially as the execution layer in always-on agents that call tools and validate results. Another documented use case is security alert monitoring, where it can enrich alerts, classify incidents, query logs, and prepare findings. It is also described as useful for answering billing questions in high-volume agentic workflows.
import requests r = requests.post( "https://api.aimlapi.com/v1/chat/completions", headers={"Authorization": "Bearer " + AIMLAPI_KEY}, json={ "model": "nvidia/nemotron-3.5-lightning", "messages": [ { "role": "user", "content": "Hello!" } ] }, ) print(r.json())
const r = await fetch("https://api.aimlapi.com/v1/chat/completions", { method: "POST", headers: { Authorization: `Bearer ${process.env.AIMLAPI_KEY}`, "Content-Type": "application/json", }, body: JSON.stringify({ "model": "nvidia/nemotron-3.5-lightning", "messages": [ { "role": "user", "content": "Hello!" } ] }), }); console.log(await r.json());
curl -X POST https://api.aimlapi.com/v1/chat/completions \ -H "Authorization: Bearer $AIMLAPI_KEY" \ -H "Content-Type: application/json" \ -d '{"model":"nvidia/nemotron-3.5-lightning","messages":[{"role":"user","content":"Hello!"}]}'
OpenAI-compatible — swap the base URL and it works with your existing SDK.
| Type | Price |
|---|---|
| Input | |
| Output | |
This model is currently free to use.
| Benchmark | Score | What it measures | Source | Retrieved |
|---|---|---|---|---|
| Intelligence | 13.6 | Composite score across standardised reasoning, knowledge and problem-solving evaluations, measured independently by Artificial Analysis | Source | September 12, 2026 |
| Coding | 26.8 | Composite score across standardised coding evaluations, measured independently by Artificial Analysis | Source | September 12, 2026 |
| SWE-bench Verified | 51.56 | Resolving verified real GitHub issues | Source | September 21, 2026 |
| Terminal-Bench | 24.58 | Autonomous shell/terminal task completion | Source | September 21, 2026 |
| MMLU-Pro | 81.94 | Multi-discipline knowledge + reasoning (harder MMLU) | Source | September 21, 2026 |
| GPQA Diamond | 75.44 | Google-proof graduate science questions (hardest subset) | Source | September 21, 2026 |
| Humanity's Last Exam | 11.72 | Expert-level questions across many domains | Source | September 21, 2026 |
| Model | Input | Output | Context | Best for |
|---|---|---|---|---|
Nemotron 3.5 Lightning This page | High-throughput agentic workloads | |||
| Long-context, multimodal & agentic workflows | ||||
| Reasoning + agents | ||||
| Complex reasoning, coding and agentic workflows | ||||
| Fast, low-cost reasoning and tool use at scale |
Nemotron 3.5 Lightning has a 1,000,000 tokens context window and can return up to 65,536 tokens.
Nemotron 3.5 Lightning takes text as input and returns text.
Use nvidia/nemotron-3.5-lightning as the model id. Requests go to https://api.aimlapi.com/v1/chat/completions.
Nemotron 3.5 Lightning became available on August 11, 2026.
Nemotron 3.5 Lightning is priced at input $0 / 1M, output $0 / 1M.
Yes, Nemotron 3.5 Lightning can stream responses as they are generated.
Nemotron 3.5 Lightning was built by NVIDIA.
It is a mixture-of-experts model with 30B total parameters, using 3B active parameters per task.
