NVIDIA's Nemotron 3.5 Lightning trims MoE for speed
A 30-billion-parameter mixture-of-experts model activates just 3 billion parameters per token, using a hybrid Mamba design to keep inference fast.

NVIDIA has published Nemotron-3.5-Lightning-30B-A3B, a new open-weights language model that leans on a mixture-of-experts (MoE) design to balance capability against inference cost. The model carries 30 billion total parameters but activates only about 3 billion per token, a ratio that lets it punch above the compute budget of a dense model its size. It is available now on Hugging Face.
The "Lightning" branding points to the model's focus on throughput. Alongside the sparse MoE routing, NVIDIA uses a hybrid Mamba architecture, blending state-space layers with attention to reduce the quadratic cost that traditional transformers pay on long sequences. The result is a model built for faster generation without discarding reasoning ability, which is listed among its core modalities.
Why it matters
Efficiency-first releases are becoming NVIDIA's signature in the open model space, and this one packages several of those ideas together:
- A 30B/3B-active MoE that keeps per-token compute low
- A hybrid Mamba backbone for cheaper long-context handling
- NVFP4 quantization, tuned for NVIDIA's own hardware, to shrink the memory footprint
- Multilingual coverage rather than English-only training
The NVFP4 format is the notable detail here. As a 4-bit floating-point scheme aligned with NVIDIA's latest accelerators, it signals that the company is optimizing not just for accuracy benchmarks but for how these models actually run on its chips. For teams weighing self-hosted reasoning models, that combination of sparse activation and aggressive quantization is the pitch.
NVIDIA has not published a context length for this variant, and the model ships under a custom license rather than a standard open-source one, so prospective users should review the terms before building on it.
Sources
- Visit
nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4
Hugging Face
More from NVIDIA
All NVIDIA releases →NVIDIA's Cosmos 3 Brings Open Interactive World Models
The new open release pushes beyond passive video generation toward simulated worlds that respond to input.
NVIDIA Releases Cosmos 3 for World Simulation
The new open models push NVIDIA's world-foundation effort toward interactive, real-time video generation.

NVIDIA releases compact Dory 1.0 4B text model
The new Nemotron Labs model packs a 4-billion-parameter language model trained on NVIDIA's Nemotron datasets.
More in Text / LLM
All Text / LLM →
Microsoft's Aion 1.0 Instruct is a compact 2.3B chat model
A small, multilingual instruction-tuned model aimed at chat and summarization lands on Hugging Face.
SenseTime Releases Looped 9B Multimodal Model
The company's new SenseNova-Loop-9B pairs vision-language capabilities with a recurrent looped-transformer design.

Tencent releases Hunyuan Hy4-flash multimodal MoE
The lightweight, mixture-of-experts model adds vision and text instruction-following to Tencent's open Hunyuan lineup.
0 comments
No comments yet. Be the first to weigh in.