Nemotron 3.5 Lightning 30B A3B

NVIDIAMCP Tool CallReasoning

nvidia/nemotron-3.5-lightningOpen Weights

Fast NVIDIA Nemotron MoE for reliable agentic tasks across enterprise workloads

Benchmark Pricing$0.05 / 1M in$0.20 / 1M out$0.01 cache read

Context: 262,144 tokens•Max Output: 262,144 tokens•Knowledge: 2025-09•Released: 2026-08-11

TECHNICAL SPECIFICATIONS

Context Window262,144262.1K tokens

Max Output262,144262.1K tokens

LicenseOpen SourceSelf-hostable

Input Modalities

Text

Function / Tool CallingSupported (MCP Ready)

Structured OutputsGuaranteed Schema

MODEL CONTEXT PROTOCOL (MCP) READINESS

✓ Nemotron 3.5 Lightning 30B A3B natively supports OpenAI / Anthropic format tool calling and is fully compatible with MCP servers and agent loops.

MCP Context Window Recommendations

With a context capacity of 262,144 tokens, this model can handle moderately large MCP tool definitions and multi-turn conversation logs comfortably.

STANDARDIZED BENCHMARKS & EVALS

MMLU-Proscore

81.94

AA-Omnisciencescore

17.5

SciCodescore

32.6

PinchBenchscore

85.37

BrowseCompscore

36.97

AA-LCRscore

GPQA Diamondaccuracy

75.44

Humanity's Last Examaccuracy

11.72

SWE-Bench Verifiedaccuracy

51.56

SWE-Bench Multilingualaccuracy

39.33

Terminal-Benchaccuracy

24.58

tau3-benchscore

9.28

GDPval-AAElo

832

IFBenchaccuracy

71.88

API CALL INVOCATION EXAMPLE

// Example: Calling Nemotron 3.5 Lightning 30B A3B via OpenAI SDK
import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.NVIDIA_API_KEY,
});

const response = await client.chat.completions.create({
  model: "nvidia/nemotron-3.5-lightning",
  messages: [{ role: "user", content: "Hello! Can you inspect MCP tools?" }],
});

console.log(response.choices[0].message.content);