Nemotron 3.5 Lightning 30B A3B
Active Parameters
30B
Context Length
262K
Modality
Text
Architecture
Mixture of Experts
KV Cache Rate
52 KB/tok
License
other
Release Date
1 Aug 2026
Knowledge Cutoff
-
API Pricing (per 1M)
Input: $0.06 · Output: $0.20
Hardware & System Requirements
Precision
Context
Workload Memory: 18.27 GB
INT4 · 8K context
Base Weights: 15.37 GB
KV Cache: 0.41 GB
Runtime & Graphs: 2.50 GB
Minimum Required Hardware
1x NVIDIA L4 24GB
18.3 / 21.6 GB usable
2x RTX 4090 24GB
18.3 / 43.2 GB usable · TP 2
Apple M3 Pro 36GB
18.3 / 27.0 GB usable
Determine requirements for your specific workload:
Open in VRAM Calculator
Runtime requirements
Checkpoint format
FP16
CUDA
>= 11.8
PyTorch
>= 2.1
Transformers
>= 4.40.1
| Architecture | Precision | Attention kernel |
|---|---|---|
Blackwell, Hopper | Native FP16 | FlashAttention-3 |
Ada Lovelace, Ampere | Native FP16 | FlashAttention-2 |
Architecture Diagram
Evaluation Benchmarks
Rank
#231
| Benchmark | Score | Rank |
|---|---|---|
Professional Knowledge | 0.819 | 23 |
Agentic Coding | 0.20 | 56 |
Mathematics | 0.67 | 58 |
General | 0.44 | 64 |
Data Analysis | 0.43 | 64 |
LiveBench Average | 0.44 | 64 |
Coding | 0.46 | 65 |
Reasoning | 0.43 | 66 |
Coding Index | 0.27 | 122 |
General Text | 1351 | 147 |
Rankings
Overall Rank
#231
Coding Rank
#154
About Nemotron 3.5 Lightning 30B A3B
Nemotron 3.5 Lightning is an open mixture-of-experts model from NVIDIA activating 3B parameters out of 30B total. It is optimized for high-throughput agentic workflows and efficient long-context reasoning.
Specifications
Attention
Attention Structure
Grouped-Query Attention
Attention Heads
32
Key-Value Heads
2
Attention Head Dimension
128
Position Embedding
ROPE
RoPE Theta
10,000
Linear Attention
Yes
Linear Attention Ratio
46.2%
Normalization
RMS Normalization
Activation Function
ReLU2
Dimensions
Base Precision
FP16
Hidden Dimension Size
2,688
Number of Layers
52
FFN Intermediate Size (Dense)
1,856
Multi-Token Prediction Heads
1
Tokenizer
Vocabulary Size
131,072
Mixture of Experts
Total Expert Parameters
3.0B
Number of Experts
128
Active Experts
6
Shared Experts
1
FFN Intermediate Size (per Expert)
1,856
Resources
About Nemotron 3.5
NVIDIA's Nemotron 3.5 family of open models delivers leading efficiency and accuracy for agentic AI applications. Built on an advanced hybrid Mamba-2 SSM and sparse Mixture-of-Experts architecture with native Multi-Token Prediction (MTP) and speculative decoding harnesses, Nemotron 3.5 models are purpose-built for the high-volume execution layer of always-on agents.