NVIDIA Releases Nemotron 3.5 Lightning 30B-A3B: 3B-Active MoE Model With 1M-Token Context, Quantized for Single-GPU Depl
NVIDIA has published NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4, a 30-billion-parameter Mixture-of-Experts model with only 3B active parameters, a hybrid Mamba-2/MoE/Attention architecture, and support for up to 1 million tokens of context. The NVFP4-quantized checkpoint is designed to run on a single DGX Spark (GB10) or H100 GPU.
NVIDIA Nemotron 3.5 Lightning 30B-A3B-NVFP4 — Quick Specs
What was released
NVIDIA has published NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4 on Hugging Face, an open-weight large language model with 30 billion total parameters but only 3 billion active per forward pass, thanks to a Mixture-of-Experts (MoE) design. The architecture interleaves Mamba-2 state-space layers, MoE layers, and select Attention layers — a hybrid approach NVIDIA is positioning for long-running agentic workloads.
The model supports a context window of up to 1,000,000 tokens and ships pre-quantized in NVFP4 format for efficient single-GPU inference. According to NVIDIA, it can run on one DGX Spark (GB10) unit or a single H100 GPU, with broader hardware support spanning Blackwell (GB200, GeForce RTX 5090), Hopper (H100, H200), and Ampere via W4A16.
The listing shows a release date of August 11, 2026, with a pre-training data cutoff of September 2025 and post-training data cutoff of May 2026. NVIDIA says the model was pre-trained on more than 20 trillion tokens using an NVFP4 recipe and Megatron-LM, then underwent continued pre-training for Multi-Token Prediction (MTP), supervised fine-tuning, and reinforcement learning via GRPO using NeMo RL and NeMo Gym. A final post-training quantization stage ("Four Over Six NVFP4") produced the NVFP4 checkpoint evaluated here.
Benchmark results
NVIDIA published side-by-side scores comparing the BF16 and NVFP4 checkpoints, measured under its internal NeMo Gym / Nemo Evaluator SDK harness — these are NVIDIA's own reported numbers, not independently verified:
- MMLU Pro: 81.94 (BF16) vs. 81.62 (NVFP4)
- GPQA Diamond: 75.44 vs. 75.57
- SWE-bench Verified: 51.56 vs. 52.80
- SWE-bench Multilingual: 39.33 vs. 36.47
- Terminal-Bench 2.1: 24.58 vs. 23.46
- HLE (text-only): 11.72 vs. 10.47
- IFBench (loose): 71.88 vs. 72.88
- AA-LCR (long context): 52.00 vs. 49.19
The quantized checkpoint tracks closely with the BF16 version across most tasks, with modest drops on long-context (AA-LCR) and multilingual coding benchmarks, and slight gains on SWE-bench Verified and IFBench.
Deployment details
NVIDIA recommends a sampling temperature of 1.0 and top_p of 0.95. The model integrates with vLLM (nightly build v0.27.1) and supports speculative decoding via a companion "DSpark" checkpoint, alongside MTP and DFlash acceleration methods. It supports English, Spanish, French, German, Italian, and Japanese, plus coding languages, and is released under NVIDIA's OpenMDW-1.1 license for commercial use.
What this means
This release continues NVIDIA's push toward efficient, low-active-parameter MoE models that can run on a single consumer or workstation-class GPU rather than multi-node clusters. A 3B-active/30B-total design with 1M-token context and native NVFP4 quantization targets a specific niche: long-running autonomous agents and sub-agent workflows that need large context without the memory footprint of a dense model. The benchmark parity between BF16 and NVFP4 checkpoints — NVIDIA's own numbers — suggests the quantization recipe preserves most capability while cutting deployment cost. Independent verification of these scores has not yet appeared, and pricing is not applicable since this is an open-weight release rather than an API product.
Related Articles
Nvidia Releases Nemotron 3.5 Lightning: A 31.6B-Parameter Open Model Built for Speed, Not Peak Intelligence
Nvidia's Nemotron 3.5 Lightning, a 31.6B-parameter open-weight model with only 3.6B active parameters, matches OpenAI's gpt-oss-120b on the Artificial Analysis Intelligence Index while delivering the fastest throughput in its class at nearly 670 tokens per second. The model posts especially large gains on agentic benchmarks, beating both gpt-oss-120b and the larger Nemotron 3 Super.
NVIDIA Releases Alpamayo 2 Super, a 34B Vision-Language-Action Model for Autonomous Driving
NVIDIA has released Alpamayo 2 Super, a 34B-parameter foundation model for autonomous vehicle development that combines a 32B vision-language backbone with a 2.3B-parameter diffusion action decoder. The model handles trajectory prediction, visual question answering, 2D grounding, and auto-labeling, and posts a Lingo-Judge score of 79.2 on LingoQA reasoning evaluation.
NVIDIA Releases Magpie TTS Multilingual Update: 364M-Parameter Open-Weights Model Now Supports 12 Languages, Sub-50ms La
NVIDIA's Magpie TTS Multilingual, a 364M-parameter open-weights text-to-speech model, now supports 12 languages after adding Modern Standard Arabic, Korean, and Brazilian Portuguese. The model achieves 32ms time-to-first-audio on B200 GPUs and improves speech quality across French, Spanish, and German.
Unsloth Releases GGUF Quantizations of Meta's Muse Glimmer 30B Agentic Model
Unsloth has published GGUF quantizations of Muse Glimmer-30B, a dense 29.6B-parameter causal transformer with a dedicated perception encoder, attributed to Meta Superintelligence Lab in the model card. The model targets autonomous agentic tasks on consumer hardware with a 131,072-token context window and 4-bit quantization under 20GB.
Comments
Loading...