NVIDIA 发布 Nemotron 3.5 Lightning，专为长时运行智能体设计，30B 总参数/3B 激活参数，支持 1M token 上下文，可商用。支持单卡部署（DGX Spark 或 H100），并加入多 token 预测与 NVFP4 量化，输出速度最高提升 4 倍。PinchBench 上准确率 86%，完成速度比 Qwen3.6 35B 快 30%。
NVIDIA released Nemotron 3.5 Lightning for long-running agent execution.
- 30B total / 3B active parameters, with up to 1M tokens of context.
Ready for commercial use.
- Local deployment: NVIDIA ships BF16 and much smaller NVFP4 checkpoints, lists 1× DGX Spark or 1× H100 for single-GPU deployment, and also lists RTX 5090 among supported hardware.
- NVIDIA then adds multi-token prediction, DSpark and DFlash speculative decoding, plus NVFP4 quantization to push generation speed further.
- NVIDIA claims up to 4× output speed; on PinchBench, 86% accuracy and 30% faster completion than Qwen3.6 35B at similar accuracy.
Lightning strikes for continuous and long-run agents! Nemotron 3.5 Lightning is smart, fast, efficient and open.在 X 查看被引用的帖子
来源：Rohan Paul · x.com