Oct 20, 2026
媒体
TechCrunch: Nvidia launches new open models to run agents in the enterprise and on PCs
TechCrunch:Nvidia launches new open models to run agents in the enterprise and on PCs(2026 年 10 月 19 日)
TechCrunch,Julie Bort(该媒体的 Venture Editor),署名「8:37 AM PDT · October 19, 2026」(article:published_time meta 2026-10-19T15:37:09+00:00,JSON-LD datePublished 一致,16:25:21 修改;本次经原始 HTML 打开并完整读取)。事件锚点:「Nvidia on Thursday launched new open-weight models designed for always-on agents and edge computing.(英伟达于周四发布了为常驻 agent 与边缘计算设计的新开放权重模型。)」规模:「The Nemotron 3.5 Super VL is a large, 120-billion-parameter model geared for agents used by the enterprise.(Nemotron 3.5 Super VL 是一个大型、1200 亿参数的模型,面向企业使用的 agent。)」;「While it uses fewer parameters than the Nemotron 3 Ultra model, Nvidia says it can deliver similar accuracy in instruction-following and tool-use tasks.(虽然参数量少于 Nemotron 3 Ultra,但英伟达称其在指令遵循与工具使用任务上可达到相近准确率。)」PC 层:Bolt 模型「are intended to run locally on the new AI PCs (running the Nvidia RTX Spark chip) and workstations running the DGX Spark. Such PCs, aimed at consumers, programmers, and enterprise knowledge workers, are being released right now into the market through partners like Microsoft and Dell.(按设计在新 AI PC——搭载 Nvidia RTX Spark 芯片——与运行 DGX Spark 的工作站上本地运行。这类 PC 面向消费者、程序员与企业知识工作者,正通过 Microsoft、Dell 等合作伙伴进入市场。)」Briski 引语,按 TechCrunch 的转写(其职务写作「vice president of Open Models and Software(开放模型与软件副总裁)」):「Any organization that wants to run agents on Nvidia infrastructure now has an open, optimized model for their tier without having to choose between capability and deployability,(任何想在英伟达基础设施上运行 agent 的组织,现在都有一为其层级开放且优化过的模型,无需在能力与可部署性之间取舍。)」说法框架:「Nvidia claims that the Super VL completes tasks faster than competing models from Anthropic and OpenAI (Opus 5, GPT 6.1 Sol, Luna, Astra) when post-trained on specific agentic tasks. These tasks include document intelligence, financial accounting, scientific experimentation, medical insurance claims, and security attack resilience.(英伟达声称,在特定 agentic 任务上做后训练后,Super VL 完成任务的速度超过 Anthropic 与 OpenAI 的竞品模型——Opus 5、GPT 6.1 Sol、Luna、Astra。任务包括文档智能、财务会计、科学实验、医疗保险理赔、安全攻击韧性。)」采用方:「a list of heavy-hitter companies that it says are using Super VL. These include Skydance, EY, Ramp, Crowdstrike, Perplexity, and Salesforce.(一份它称正在使用 Super VL 的重量级公司名单:Skydance、EY、Ramp、Crowdstrike、Perplexity、Salesforce。)」开放性引语,按 TechCrunch 的转写:「Unlike closed frontier models, organizations can post-train Super VL on their own data, deploy it on their own infrastructure and own what they build,(与封闭前沿模型不同,组织可用自己的数据后训练 Super VL、部署在自己的基础设施上,并拥有自己构建的东西。)」Bolt 细节:「a 74-billion parameter model designed for local agentic tasks ... to handle multistep reasoning, tool calling, and web browsing. The smallest Bolt E4B is aimed at conversational agents.(一个为本地 agentic 任务设计的 740 亿参数模型……处理多步推理、工具调用与网页浏览。最小的 Bolt E4B 面向对话 agent。)」可用性:「Super VL and Bolt 74B-A9B are available now, with Bolt E4B coming later this month, Nvidia says.(Super VL 与 Bolt 74B-A9B 现已可用,E4B 将于本月晚些时候到来——英伟达称。)」与官方博客的转写差异,各归各、绝不熔接:「own what they build(拥有自己构建的东西)」(TechCrunch)对官方的「oversee what they build(监督自己构建的东西)」;Briski 引语 without 一词前无逗号(TechCrunch)对官方有逗号;TechCrunch 写本月晚些时候,官方写 10 月 23 日;点名竞品(TechCrunch)对官方不点名的最高 5 倍(Artificial Analysis 指数)。