NVIDIA has released Nemotron 3.5 Lightning, a new open AI model designed to make autonomous AI agents faster, more efficient, and easier to deploy on local hardware.

Released August 11, 2026, the model uses a 30-billion-parameter mixture-of-experts architecture with only 3 billion active parameters, reducing the amount of compute required for each task.

NVIDIA is positioning Lightning as an execution-focused model for AI agents handling frequent operations such as tool calls, code review, result validation, sub-agent tasks, security monitoring, and other high-volume workloads.

Nemotron 3.5 Lightning combines Mamba-2, mixture-of-experts layers, and selected attention layers while supporting a context window of up to 1 million tokens. NVIDIA also includes NVFP4 and BF16 checkpoints, Multi-Token Prediction, DSpark and DFlash speculative-decoding methods to increase inference performance.

According to NVIDIA’s testing, Lightning can deliver up to 4x the output speed of comparable open models and completed 10,000 PinchBench agentic tasks about 30% faster than Qwen3.6 35B at similar accuracy.

The model can be deployed through common local AI tools including Ollama, LM Studio, llama.cpp and Unsloth, giving developers several paths for running it outside a traditional cloud AI service.

The release is particularly significant for the growing local AI and agentic AI ecosystem. NVIDIA lists single-GPU deployment on the DGX Spark GB10 and H100, while supported hardware also includes Blackwell systems such as the GeForce RTX 5090.

NVIDIA has released the model under its OpenMDW 1.1 license, with model weights, training data and recipes available for developers to customize, fine-tune and deploy.

Nemotron 3.5 Lightning also arrives alongside NeMo Switchyard, NVIDIA’s model-routing library designed to send complicated planning jobs to frontier models while directing routine execution toward smaller, cheaper models such as Lightning.

Together, the releases highlight a growing AI architecture in which developers use a coordinated system of specialized models instead of relying on one massive model for every task.

^{BitcoinVersus.Tech Editor’s Note:}

^{We volunteer daily to ensure the credibility of the information on this platform is Verifiably True. If you would like to support to help further secure the integrity of our research initiatives, please donate here: 3C9o19EH5HSiwEPyCTmEKzxhNCbo2X6TTb}

BitcoinVersus.tech is not a financial advisor. This media platform reports on financial subjects purely for informational purposes.

‪Use the #promocode bi