Not established from the available sources.
NVIDIA · MODEL RELEASE TRACKER
Nemotron 3.5 Lightning
Nemotron 3.5 Lightning is a named variant in NVIDIA's Nemotron 3.5 lineup. Public reporting states Lightning was released in August and is positioned for high-throughput inference; like other Nemotron variants it is described as optimized for NVIDIA hardware.CURRENT SNAPSHOT2/5 DIMENSIONS WITH DATA
The dimensions that change the decision.
Not established from the available sources.
Not established from the available sources.
PinchBench accuracy and task speedReported 86% accuracy on PinchBench while completing tasks 30% faster than comparable models.
Interactive evaluation availabilityAvailable to try interactively on build.nvidia.com.
RELEASE TIMELINE
Published, source-backed release events only.
VERIFIABLE FACTS
Every value stays attached to a source and date.
BENCHMARKS · PinchBench accuracy and task speedDEVELOPER CLAIM
Reported 86% accuracy on PinchBench while completing tasks 30% faster than comparable models.
CAPABILITIES · Mixture-of-Experts parameter activationDEVELOPER CLAIM
Uses a Mixture-of-Experts architecture that activates 3B of its 30B total parameters per token.
CAPABILITIES · Throughput relative to dense modelsDEVELOPER CLAIM
At equal total parameter counts, achieves higher token throughput than comparable dense models (e.g., Gemma 4 31B), though latency advantage narrows at high concurrency.
CAPABILITIES · Total parameter countDEVELOPER CLAIM
Total parameter count reported as 30B.
AVAILABILITY · Interactive evaluation availabilityDEVELOPER CLAIM
Available to try interactively on build.nvidia.com.
LIMITATIONS · Fine-tuning and quantization caveatsDEVELOPER CLAIM
Fine-tuning MoE models requires care to avoid router imbalance; quantization affects router and recurrent-projection layers differently than dense-model components.
RELEASE · Release timingDEVELOPER CLAIM
Reported released in August.
CAPABILITIES · Designed for high throughputDEVELOPER CLAIM
Described as built for high throughput (positioned for high-throughput inference).
CAPABILITIES · Hardware optimizationDEVELOPER CLAIM
Reportedly optimized for NVIDIA hardware, like other Nemotron variants.
WHAT CHANGED
Stored passport versions, without reconstructed history.
CAPABILITIES · Hardware optimization+ Reportedly optimized for NVIDIA hardware, like other Nemotron variants.
CAPABILITIES · Designed for high throughput+ Described as built for high throughput (positioned for high-throughput inference).
RELEASE · Release timing+ Reported released in August.