NVIDIA 开源 Nemotron 3.5 Super VL：120B MoE 多模态智能体模型，每 token 仅激活 12B
https://t.co/GyQi40zsgr
NVIDIA 开源了一个 120B 多模态智能体模型，每 token 只激活 12B，跑智能体任务比大 5 倍的 Ultra 还快，ms-swift 里就能微调。
NVIDIA 开源发布 Nemotron 3.5 Super VL，提供 BF16 和 NVFP4 两种格式。该模型为 120B MoE 多模态架构，每个 token 只激活 12B 参数。在 Hermes Bench 和 LangChain 评测中，其智能体表现超过参数量大 5 倍的 Nemotron 3 Ultra，同尺寸级别吞吐量最高提升 5 倍，评测的智能体任务完成速度最高快 4 倍。ModelScope 的 ms-swift 已支持该模型的推理与 LoRA 微调，NVIDIA NeMo 可用于 SFT 和 RL 定制。部署方面支持 NVFP4 QAD、多 token 预测与跨轮上下文复用，兼容 vLLM、SGLang、TRT-LLM 和 NIM。
https://t.co/GyQi40zsgr
x.com/ModelScope2022… ModelScope @ModelScope2022 NVIDIA Nemotron 3.5 Super VL is open released in BF16 and NVFP4: a 120B MoE multimodal agent model that activates just 12B parameters per token. ModelScope’s ms-swift supports both inference and fine-tuning, helping developers put the model to work and adapt it to specialized multimodal applications. 🤖 modelscope.ai/collections/nv… 8 � github.com/modelscope/ms-… Ms 🏆 Agent performance: Outperforms the 5x larger Nemotron 3 Ultra on the reported Hermes Bench and LangChain evaluations. ⚡ Faster workflows: Delivers up to 5x higher throughput in its size class and completes evaluated agent tasks up to 4x faster. 👁️ Multimodal reasoning: Understands text, images, charts, documents, and video while reasoning, coding, and using tools. 🧰 Built to customize: Fine-tune with your own multimodal datasets using LoRA in ms-swift, or use NVIDIA NeMo for SFT and RL to adapt the model to domain data, tools, policies, and workflows. 🧠 Efficient deployment: NVFP4 with QAD, multi-token prediction, and cross-turn context reuse reduce repeated inference work. Supports vLLM, SGLang, TRT-LLM, and NIM across local, data center, and cloud environments. 🔗 View Quoted Tweet 💬 0 🔄 0 ❤️ 4 👀 1102 📊 1 ⚡
