Sakhanda Wire
NVDA $153.59 -1.64% MSFT $507.00 -2.12% GOOGL $279.98 -1.32% META $619.94 -1.03% AMZN $256.62 -0.44%
← Back to the news

NVIDIA launches Nemotron 3.5 Lightning and Switchyard router

NVIDIA has introduced two open technologies for always-on AI agents: Nemotron 3.5 Lightning and the NeMo Switchyard routing library. Lightning is a 30B mixture-of-experts model with 3B active parameters, designed for high-volume execution tasks such as tool calls, result validation and subagent delegation. Its hybrid Mamba-2, MoE and Attention architecture supports a 1M-token context window.

The model was pretrained on more than 20 trillion tokens and is available with open weights, training data and recipes under the permissive OpenMDW-1.1 license. NVIDIA says it supports commercial use, can run on one DGX Spark with GB10 or one H100, and delivers up to four times the output speed of similarly sized models. On 10,000 PinchBench tasks, NVIDIA reports 86% accuracy and completion 30% faster than Qwen3.6 35B at comparable accuracy. NVFP4 and BF16 checkpoints are provided, alongside DSpark and DFlash draft models for speculative decoding.

Switchyard routes individual workflow steps to models based on capability and efficiency. In a LangChain benchmark of 145 multi-turn tasks, its escalation router reduced cost by 74% compared with a frontier-only setup, sending 7% of calls to the frontier model, with an approximately six-point accuracy trade-off. The library supports OpenAI, Anthropic and Responses API requests.

Written by Sakhanda Wire AI from reporting by MarkTechPost. Facts, quotes and full context are in the original article below.

Originally published by MarkTechPost on

Read the original on MarkTechPost ↗

The reporting and the image belong to MarkTechPost and are credited and linked above; the text on this page is Sakhanda Wire's own summary.

← Back to the news

More stories

All the latest news