Nemotron 3.5 LightningOpen Source
Released
Nemotron 3.5 Lightning is an AI model released by NVIDIA on Tuesday, Aug 11 2026, 68 days after Nemotron 3 Ultra. It is an open-source model — weights, and training code/data, released under an open license. It is a 30B parameter model with a 1M token context window. Benchmark results (shown below) cover SWE-Bench Verified, and GPQA Diamond.
Benchmarks
Coding
Reasoning & science
About
Nemotron 3.5 Lightning, released August 11, 2026, was a mid-generation refresh of the small tier rather than a new family: 31.6 billion total parameters with 3.6 billion active, tuned specifically for long-running agentic work. NVIDIA shipped it alongside NeMo Switchyard, an open-source router for agent systems that mix models of different sizes — a pairing that made the positioning explicit, with Lightning as the fast default a router falls back from only when a task needs a larger model.
At launch it reported 75.4% on GPQA Diamond and 51.6% on SWE-Bench Verified — close to gpt-oss-120b-class accuracy at roughly a quarter of the total parameters — with NVIDIA claiming up to four times the output speed of comparably sized models. The release went out under the permissive OpenMDW-1.1 license with weights, training data and recipes included, the fullest openness of any Nemotron release to that point.
Compare Nemotron 3.5 Lightning with
Suggested comparisons
Frequently asked questions
Nemotron 3.5 Lightning was released by NVIDIA on Tuesday, Aug 11 2026.
Nemotron 3.5 Lightning was built by NVIDIA. The dominant AI chipmaker, publishing the open Nemotron model family. Ships weights, datasets and training recipes to seed demand for its hardware rather than to sell a chatbot.
Nemotron 3.5 Lightning reports 2 tracked benchmark scores — SWE-Bench Verified: 51.6%; GPQA Diamond: 75.4%. Scores are the figures published at release by NVIDIA.
Nemotron 3.5 Lightning has a context window of 1M. That is the maximum amount of input plus output the model can hold in a single request.
Nemotron 3.5 Lightning is reported at 30B parameters.
Yes. Nemotron 3.5 Lightning is open source: the model weights are published together with training code and data under an open license, so it can be downloaded, self-hosted, fine-tuned, and in principle reproduced from scratch.
NVIDIA's previous tracked release was Nemotron 3 Ultra on Jun 4 2026, 68 days earlier. It is the most recent NVIDIA model tracked on AI Release Tracker.