Nvidia's Nemotron 3.5 Super VL outperforms Claude and GPT models

EY: EY is a professional services firm offering audit, tax, consulting, and advisory services. It has developed a merger-and-acquisition due-diligence agent using the customized Super VL model. R1: R1 is an organization engaged in AI model adaptation. It is among the entities post-training Super VL. AI21: AI21 is an AI company focused on enterprise language models and applications. It participated in post-training Super VL for cybersecurity use cases alongside Alice. Ramp: Ramp offers corporate cards and spend management software. It is listed among companies post-training Super VL for internal tasks. Alice: Alice is an AI firm working on cybersecurity solutions. It contributed to customizing Super VL for cybersecurity tasks with AI21. Dream: Dream is an AI company involved in model customization. It is named as one of the organizations adapting Super VL. Hebbia: Hebbia develops AI platforms for knowledge work and research. It deployed a customized Super VL model for credit-agreement analysis via AfterQuery. Mercor: Mercor provides AI tools for recruiting and talent assessment. It is among the organizations customizing Super VL for coding agents. NVIDIA: NVIDIA develops graphics processing units, AI software platforms, and enterprise AI solutions focused on accelerating computing workloads. The company is releasing multiple open Nemotron models tailored for multimodal enterprise agents that combine language, vision, and tool use while supporting post-training and flexible deployment. Scale AI: Scale AI supplies data labeling, evaluation, and post-training services for AI models. It contributed to customizing Super VL for Skydance’s vendor-credit finance agent. Skydance: Skydance is a media and entertainment company. It is using a post-trained version of Super VL for a vendor-credit finance agent through work with Scale AI. Synopsys: Synopsys provides electronic design automation and semiconductor IP solutions. It is integrating the model into its tools. Atlassian: Atlassian develops collaboration and productivity software including Jira and Confluence. It is integrating Super VL into its platform offerings. AfterQuery: AfterQuery is a company specializing in AI applications for financial analysis. It worked with Hebbia on credit-agreement analysis using the post-trained model. ModelScope: ModelScope is a platform for discovering and using machine learning models. It is listed as a distribution channel for the released models. NVIDIA NIM: NVIDIA NIM delivers optimized inference microservices for deploying AI models. Organizations can download the models as NIM microservices from the NGC catalog. OpenRouter: OpenRouter provides a unified API for accessing multiple AI models. It serves as one of the launch access routes for the new Nemotron models. Perplexity: Perplexity operates an AI-powered search and research platform. It is integrating the model into its services. Salesforce: Salesforce delivers customer relationship management and enterprise software platforms. It is adding Super VL capabilities to its offerings. CrowdStrike: CrowdStrike provides cloud-native cybersecurity solutions. It is incorporating the model into its security products. Hugging Face: Hugging Face operates an open-source platform for hosting, sharing, and deploying machine learning models. Super VL weights are available there, and NVIDIA has agreed to acquire the company. Lila Sciences: Lila Sciences works on scientific AI applications. It is customizing Super VL for its workflows. Nemotron 3 Ultra: Nemotron 3 Ultra is NVIDIA's larger model positioned for demanding agentic reasoning workloads. It appears in the timeline of releases leading to the current Super VL addition of visual capabilities. Nemotron Bolt E4B: Nemotron Bolt E4B is NVIDIA's compact open model aimed at conversational agents, voice interfaces, and embedded systems with tool use. It is scheduled for release to support local agent workloads on edge hardware. Dassault Systèmes: Dassault Systèmes creates 3D design and simulation software for industries. It is integrating Super VL into its enterprise solutions. Nemotron 3 Nano Omni: Nemotron 3 Nano Omni is an earlier NVIDIA multimodal model that integrates text, images, audio, and video for document intelligence and interface reading tasks. It is referenced as part of the progression toward placing visual capabilities within the larger Super-tier lineup. Nemotron 3.5 Super VL: Nemotron 3.5 Super VL is NVIDIA's open 120-billion-parameter multimodal model designed to process PDFs, charts, tables, images, and webpages while following instructions and invoking tools. It forms the core of the announcement, showing strong gains on specialized enterprise tasks after partner-led post-training and outperforming certain closed models in those evaluations. Nemotron 3.5 Lightning: Nemotron 3.5 Lightning is an NVIDIA model targeting high-volume, repeated steps in extended agent workflows. It is cited as part of the recent evolution of the Nemotron family that Super VL builds upon for document and image understanding. Nemotron 3.5 Bolt 74B-A9B: Nemotron 3.5 Bolt 74B-A9B is NVIDIA's open model that activates a subset of parameters for efficient local execution of coding, research, and multistep agent tasks. It extends the Nemotron family with options for on-device or local deployment alongside the larger cloud-scale models. Multimodal Competition: Meta’s Llama 4 and Google’s Gemma 4 also support combined visual and textual inputs in open models. Enterprise Post-Training: Dozens of organizations across finance, legal, cybersecurity, and other sectors are adapting Super VL to their own documents, policies, and tools. Local Deployment Options: Smaller models in the Nemotron family enable agent workloads to run on local hardware without sending data to cloud services.