NVIDIA Releases Open Source Nemotron 3.5 Lightning Model Designed for Agent Tool Invocation

NVIDIA has released the open-source model Nemotron 3.5 Lightning, specifically designed for long-running Agent tasks. The model has a total of 30 billion parameters, with 3 billion activated per token, primarily responsible for high-frequency tasks such as tool invocation, result checking, and sub-Agent scheduling. NVIDIA has streamlined the division of labor for Agents, delegating complex planning to larger models like Nemotron 3 Ultra, while Lightning handles a significant amount of repetitive execution. Its output speed can reach up to four times that of comparable models, achieving an accuracy of 86% on PinchBench. In cases of similar accuracy, it completes 10,000 tasks 30% faster than Qwen3.6 35B. The model employs a MoE architecture, trained specifically for Agent Harness, and incorporates multi-token prediction and speculative decoding. Officially provided weights in BF16 and NVFP4 can run on local devices such as RTX 5090 and DGX Spark, supporting llama.cpp, Ollama, LM Studio, and Unsloth. NVIDIA has also opened up weights, some training data, and training recipes, allowing enterprises to perform post-training for tasks related to code, security, and legal matters. In showcased cases, Lightning's post-training has shown significant improvements across various specialized tasks.

This content is provided for general informational purposes only and doesn't constitute financial, investment, legal, or tax advice. Any events, rewards, online promotions, or related information mentioned herein should not be considered a recommendation, solicitation, or invitation to purchase, sell, trade, or otherwise deal in any crypto assets. Crypto assets are highly volatile and may result in loss. The availability of WEEX services, products, and related events may vary by region. You are responsible for ensuring that your participation is in accordance with applicable local laws and regulations.