NVIDIA Launches Nemotron 3.5 Lightning Optimized for Long-Running Agents
rohanpaul_ai · x · 2026-08-12
NVIDIA has released Nemotron 3.5 Lightning, a model specifically designed for continuous and long-running agent execution. It features 30B total parameters with 3B active parameters, supports up to 1M tokens of context, and is ready for commercial use.
Deployment & Optimization:
- Supports single-GPU local deployment with BF16 and much smaller NVFP4 checkpoints. Recommended hardware includes 1× DGX Spark or 1× H100, and it also lists RTX 5090 among supported hardware.
- Incorporates multi-token prediction, DSpark and DFlash speculative decoding, plus NVFP4 quantization to further push generation speed.
Performance:
- Claims up to 4× output speed.
- On PinchBench, it achieves 86% accuracy and 30% faster completion than Qwen3.6 35B at similar accuracy.
Related event: NVIDIA Unveils Nemotron 3.5 Lightning and NeMo Switchyard Router(55 posts)→
More from Infra
- Jensen Huang: Musk's AI buildout moved from K's to M's, now a CSP-tier NVIDIA customer — elonmusk · 2026-10-16
- EpochAI pegs GPT-6 training at up to 1.8e27 FLOP; scaling01 calls param estimates way too high — scaling01 · 2026-10-16
- Meta's Muse Spark Open scores 39 vs 48 for closed Spark 1.3 on AA Intelligence Index — EAccelerate_42 · 2026-10-16
- LLM spots useless 95GB-per-frame memcpy in ffmpeg dblur, cutting runtime from 2452ms to 49ms — IgorCarron · 2026-10-16
- DeepSeek-V4-Flash on 4x Radeon AI PRO R9700: the 'refused' 4-card config hits ~79 tok/s — brainchillzZ · 2026-10-16
- Tree packing lands in TRL's AsyncGRPO, cutting agentic RL training compute by over 26x — joecole · 2026-10-16