NVIDIA Nemotron 3.5 Lightning 30B-A3B
Release in the NVIDIA Nemotron family · version 3.5 Lightning (30B-A3B), GA
Nemotron 3.5 Lightning is an NVIDIA language model with 30B total and 3B active parameters, using a hybrid Mamba-2 and attention mixture-of-experts architecture with multi-token prediction and configurable reasoning. NVIDIA released full-precision BF16 weights, an NVFP4 checkpoint for deployment, draft models for speculative decoding, and a separate base checkpoint.16
- Model hub: Model card (BF16) (external site: huggingface.co)
- License: OpenMDW License Agreement, version 1.1 (external site: raw.githubusercontent.com)
- Documentation: Nemotron 3.5 Lightning training recipe (external site: github.com)
Availability and license
Overall availability
Weights are downloadable from Hugging Face under the OpenMDW-1.1 license, a copy of which is included in the model repository; the repository metadata showed no access gate when checked.123
Availability is separate from permission: read the license before using or redistributing.
OpenMDW License Agreement, version 1.1 (external site: huggingface.co)12
Apache License 2.0 (Nemotron Developer Repository training recipes) (external site: raw.githubusercontent.com)5
OpenMDW-1.1 grants free permission to use, modify, and share the model materials without restriction, including under copyright, patent, database, and trade secret rights. Redistributors must include a copy of the agreement and keep applicable origin notices. All rights end for anyone who sues claiming the materials infringe a patent or copyright, unless responding to a suit brought against them first. It imposes no conditions on outputs and disclaims warranties. The SPDX list (version 3.29.0) includes OpenMDW-1.0 but not version 1.1, so no SPDX ID is recorded. NVIDIA's Lightning recipe page states that the weights, data, and recipes are released under OpenMDW-1.1, while the Nemotron Developer Repository's only LICENSE file is Apache 2.0; this record lists the repository LICENSE for the recipe code.2465
Model-disclosure tier
Open-weight, plus published inference code, training code, and training recipe, and at least documented training-data composition.
The weights are under a license that is not on the rubric's OSI-approved list. Read its terms before use.
Public materials checklist
| Item | Status | Notes and evidence |
|---|---|---|
| WeightsCan the general public download the model parameters for this release? | Public | BF16, NVFP4, and base BF16 checkpoints are published on Hugging Face.163 |
| Inference codeIs code for running the model published? | Public | The model card gives deployment instructions for vLLM and SGLang.1 |
| Training codeIs the code used to train the model published? | Public | The Apache-2.0 Nemotron Developer Repository provides pretraining, SFT, RL, evaluation, and quantization stages for this model.65 |
| Training-data informationPublic = the training data itself can be obtained. Partial = composition or sources are documented without full access. | Partial | The model card lists the pre- and post-training datasets. Major portions are released in Hugging Face collections, some requiring access approval, while several third-party and NVIDIA datasets are listed as private and not publicly accessible.17 |
| Training recipeAre the training configuration and procedure documented in enough detail to follow? | Public | The recipe documents pretraining, SFT, RL, evaluation, and quantization stages. There is no separate technical report; NVIDIA names the model cards and recipe configs as the references, and notes the recipes use only the open-sourced data subset.6 |
| Evaluation materialsPublic = evaluation code or prompts that let others re-run the evaluations are published. Partial = results only. | Public | The model card publishes benchmark results and states that evaluation recipes and commands for reproducing them are published in NeMo Gym.1 |
What it is useful for
Run and use notes
- The model card documents serving the BF16 checkpoint with vLLM or SGLang on NVIDIA GB200, B200, and H100 GPUs, including a single H100 80GB deployment in BF16 precision at a 256K-token context, and recommends the NVFP4 checkpoint for optimized inference.1
Organization context
Provenance and derivatives
Trained by NVIDIA through pretraining, continued pretraining for multi-token prediction, supervised fine-tuning, and reinforcement learning; the base checkpoint is published separately. The model card says post-training synthetic data was generated with teacher models, including third-party open models such as GPT-OSS-120B.16
- Derived from: NVIDIA-Nemotron-3.5-Lightning-30B-A3B-Base-BF16 (external site: huggingface.co) — NVIDIA's base checkpoint, listed alongside this release in the training recipe.
Other releases in the NVIDIA Nemotron family
- NVIDIA Nemotron 3 Super 120B-A12BModel-disclosure tier (USASI rubric v0.1): Open-stack
- NVIDIA Nemotron 3 Ultra 550B-A55BModel-disclosure tier (USASI rubric v0.1): Open-weight
U.S. eligibility
Sources
This listing is not an endorsement, a safety assessment, or a federal approval.