nvidia

Nemotron 3.5 Lightning 30B A3B

nemotron

Fast NVIDIA Nemotron MoE for reliable agentic tasks across enterprise workloads

ReasoningTool CallingStructured OutputOpen Weights

Context Window

262K

Max Output

262K

Temperature

Yes

Open Weights

Yes

Knowledge Cutoff

N/A

Released

2026-08-11

Last Updated

2026-08-11

Modalities

Input:Text
→
Output:Text

Benchmarks

BenchmarkScoreMetricHarness
MMLU-Pro81.94scoreNeMo Gym / NeMo Evaluator SDK
AA-Omniscience17.5scoreNeMo Gym / NeMo Evaluator SDK
SciCode32.6scoreNeMo Gym / NeMo Evaluator SDK
PinchBench85.37scoreNeMo Gym / NeMo Evaluator SDK
BrowseComp36.97scoreNeMo Gym / NeMo Evaluator SDK
AA-LCR52scoreNeMo Gym / NeMo Evaluator SDK
GPQA Diamond75.44accuracy-
Humanity's Last Exam11.72accuracy-
SWE-Bench Verified51.56accuracyNeMo Evaluator
SWE-Bench Multilingual39.33accuracyNeMo Evaluator
Terminal-Bench24.58accuracyNeMo Evaluator
tau3-bench9.28score-
GDPval-AA832Elo-
IFBench71.88accuracy-

Available Providers (4)

ProviderInput /1MOutput /1MCache Read /1MCache Write /1MReasoningStatus
NanoGPTNanoGPTRp 895Rp 3.577Rp 179—effort—
Kilo GatewayKilo GatewayRp 448Rp 2.146Rp 224—effort—
OpenRouterOpenRouterRp 1.252Rp 3.577Rp 716—toggle—
Vercel AI GatewayVercel AI GatewayRp 895Rp 3.577Rp 179—toggle, budget_tokens—