AITOPAI热报

Nemotron 3.5 Super VL 120B A12B

NVIDIA·US·Released 2026-10-15
Open weightsMultimodalReasoningTool useVision

NVIDIA's open-weight vision-language reasoning model handling text, image, and video inputs with a 1M-token context window, designed for multimodal understanding and agentic tool use.

1Mcontext
—per 1M tokens (in / out)
—parameters
83stories in our feed

Specs

hover a value to see its source
Context window1M
Max output8K
Parameters—
Knowledge cutoff—
Licence—
Commercial useUnknown
InputText · Image · Video
OutputText
Familynemotron

API price

USD per 1M tokens

Open weights: no official API price. Self-host or use a provider.

News timeline

from the AITOP feed, one story per eventAll 83 →
2026-10-22
NVIDIA 发布面向智能体工作负载的端到端硬件栈Merve Noyan
英伟达开源检索系列模型 Nemotron 3.5 Parse、Embed 与 RerankGorden Sun
2026-10-21
Scale AI Post-Trains NVIDIA Nemotron 3.5 Super VLScale AI
Databricks expands NVIDIA collaboration for open model fine-tuningDatabricks
美国开源权重模型一周连发三款：Nemotron 3.5、Muse Spark Open、Beam 1.0Artificial Analysis
2026-10-19
Nemotron 3.5 VL Super 经皮肤科诊断后训练，与 Nolla 合作落地医疗NVIDIA AI
2026-10-17
NVIDIA publishes open midtraining recipe for agentic coding modelsDAIR.AI
Surface Laptop Ultra runs local LLMs well but runs hotBlognone
2026-10-16
FPT 与 NVIDIA 将 Nemotron-Personas-Vietnam 数据集扩大 10 倍Genk
10月16日AI快报：Meta Muse Spark Open 开源、NVIDIA 与 OpenAI 多项更新TestingCatalog
Nvidia 发布开源 Nemotron Super VL，面向企业级智能体场景Tech in Asia
2026-10-15
NVIDIA Nemotron 3.5 Super VL Gets Day-0 vLLM SupportvLLM
2026-10-13
NVIDIA Shares Nemotron Post-Training Recipe on Hugging FaceNVIDIA AI
2026-10-12
Fin 基于 NVIDIA Nemotron 微调推出 Fin Apex 2.0 客服模型NVIDIA AI
2026-10-08
黄仁勋拟再向人形机器人公司 Figure 投资 10 亿美元IT之家
2026-10-06
CoreWeave RL Rollouts Speeds Nemotron 3.5 Lightning Post-TrainingCoreWeave
2026-10-02
Policy-space scaling in reinforcement learningarXiv: DeepSeek
2026-09-29
NVIDIA runs Nemotron locally on DGXNVIDIA AI
2026-09-27
Nvidia 开源 1 亿参数说话人分离模型 Nemotron 3 DiarizationDecoder
2026-09-22
Sluicebox 用 NVIDIA Nemotron 3 Ultra 打造供应商数据智能体 LucyNVIDIA AI
2026-09-15
英伟达发布Nemotron后训练技术细节NVIDIA AI
Salesforce launches Koa推理大模型 with NvidiaIT之家
Nvidia limits use of Anthropic, OpenAI models for sensitive dataIT之家
2026-09-10
NVIDIA Nemotron 3 Embed 8B 在 Q2D-Web 基准测试中排名第一Aravind Srinivas
Nvidia and Palantir partner on sovereign AI for supply chainsIT之家
2026-09-08
Domyn Uses NVIDIA NemotronNVIDIA AI
2026-09-05
Meta的AIRA₃系统在NVIDIA竞赛中获金牌Meta AI
NVIDIA's Nemotron Model IOI 2026NVIDIA AI
2026-09-01
NVIDIA测试安全检测规则对抗新型攻击NVIDIA AI
2026-08-28
NVIDIA缩短AI模型发布周期至4-6周lmarena.ai
2026-08-23
NVIDIA专家谈教师模型与Nemotron构建lmarena.ai
2026-08-13
Nemotron 3.5 Lightning上线Fireworks，主打智能体工作流NVIDIA AI
AgileRL 与 NVIDIA 合作，Nemotron 3.5 Lightning 可在 Arena 微调NVIDIA AI
Trajectory 后训练 Nemotron 3.5 Lightning，测试成绩超过 Opus 4.6NVIDIA AI
2026-08-11
英伟达开发万亿参数开源模型 Nemotron 4，目标挑战全球顶级IT之家
NVIDIA发布Nemotron 3.5 Lightning开源MoE模型NVIDIA AI
2026-08-06
教Nemotron学希腊语：跨专业领域的RAG适配研究arXiv cs.AI
2026-07-28
在Prime Intellect上用RL训练开源模型：Nemotron Labs教程NVIDIA AI
2026-07-21
使用LangChain进行Nemotron模型微调前先调整评估框架NVIDIA AI
解读Gated Delta Networks：为何大模型参数能上T且质量好向阳乔木