Nemotron 3.5 — NVIDIA 的本地 AI 模型
作者: Jakub Rusinowski · 最后更新:
NVIDIA的智能体产品线,构建在一套混合架构上,将Mamba-2状态空间层与MoE模块交错排列,只夹杂少数注意力层。只有注意力层保留KV缓存,这正是它能在这一参数规模下宣称1M token上下文的原因。
变体
Nemotron 3.5 最小的变体在 Q4_K_M 下约需 19 GB 显存——量化权重加框架开销,不含 KV 缓存。
| 模型 | Q4 下显存 | 显存 | 上下文 | 运行 |
|---|---|---|---|---|
| Nemotron 3.5 Lightning 30B-A3B → 30B (3B active) | ~18.9 GB | 1,000,000 | ollama run nemotron-3.5-lightning:30b | |
| Nemotron 3.5 Bolt 74B → 74B (9B active) | ~22.8 GB | 262,144 | ollama run nemotron-3.5-bolt:74b-a9b |
显存为 Q4_K_M 下的量化权重加开销,与 GPU 与显存检测器使用同一引擎计算。
如何在本地运行 Nemotron 3.5
安装 Ollama,然后拉取标签。
ollama run nemotron-3.5-lightning:30b在上方选择一个尺寸,查看它自己的显存、速度估算和安装命令。
许可证
Commercial use permitted. No usage restrictions beyond attribution.
适用于: Nemotron 3.5 Lightning 30B-A3B, Nemotron 3.5 Bolt 74B推荐 GPU
目录中能在本地运行 Nemotron 3.5(至少 19 GB 显存)的最便宜 GPU 是 AMD Radeon RX 7900 XTX (24 GB).
我的 GPU 能运行 Nemotron 3.5 吗?
Nemotron 3.5 — 常见问题
How much VRAM does Nemotron 3.5 need?
Nemotron 3.5 needs about 19 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: Nemotron 3.5 Lightning 30B-A3B (19 GB, Q4_K_M); Nemotron 3.5 Bolt 74B (23 GB, IQ2_XXS QAD). On Apple Silicon, unified memory counts toward this requirement.
Can I run Nemotron 3.5 on an RTX 4090 (24 GB)?
Yes — Nemotron 3.5 runs on an RTX 4090 (24 GB) and other 24 GB cards such as the RTX 3090. Smaller variants also fit comfortably on 8–16 GB GPUs at Q4_K_M.
What quantization should I use for Nemotron 3.5?
Q4_K_M is the best balance of quality and VRAM for Nemotron 3.5 in most cases. Choose Q8_0 for near-lossless quality if you have spare VRAM, or smaller quants (Q3/Q2) only when memory is tight.
How do I run Nemotron 3.5 with Ollama?
Install Ollama, then run: ollama run nemotron-3.5-lightning:30b. This downloads Nemotron 3.5 and starts a local, OpenAI-compatible endpoint — no internet connection is needed after the initial download.