Nvidia Unveils Open AI Models: 120B-Parameter Super VL Completes Tasks 5x Faster for Enterprises
14 Sources
[1]
Nvidia launches new open models to run agents in the enterprise and on PCs
Nvidia on Thursday launched new open-weight models designed for always-on agents and edge computing. The Nemotron 3.5 Super VL is a large, 120-billion-parameter model geared for agents used by the enterprise. While it uses fewer parameters than the Nemotron 3 Ultra model, Nvidia says it can
[2]
Nvidia's new Nemotron 3.5 Super VL is another step forward for US open-weight AI - ZDNET
* Three new Nvidia open models strive for efficiency and speed. * Open models are having a renaissance on the heels of the Hugging Face hack. * US labs are catching up in the open model race. China-based AI labs have had the upper hand on competitive open-weight AI models over the last few
[3]
Nvidia woos wary Europeans with more open weights models
New embedding, parsing, and reranking models promise smarter agentic RAG workflows Nvidia hopes to capitalize on European sovereign AI demand by unloading a slew of new open weights models spanning chatbots, healthcare, and robotics. The major model devs have the resources to build highly capable
[4]
NVIDIA Open Models Advance Europe's Sovereign Intelligence Across Industries
New model families for vision and language, retrieval, tabular data, biology, robotics and quantum computing supercharge development in Europe. NVIDIA is expanding its family of open models across language, robotics, biology, and quantum computing, giving developers, researchers and enterprises
[5]
NVIDIA's latest 120B-parameter model completes tasks 5x faster
NVIDIA has unveiled three open AI models designed to bring autonomous agents to more computing environments. The lineup centers on Nemotron 3.5 Super VL, a 120-billion-parameter model built for complex tasks. Two smaller variants target users who want to run AI agents directly on local
[6]
Open From Data Center to Desktop: NVIDIA Expands Nemotron Family With Highest Efficiency Super VL and Bolt Models
5x faster, accurate and customizable Nemotron 3.5 Super VL expands NVIDIA's open models for always-on agents, with Nemotron 3.5 Bolt bringing agentic intelligence to local NVIDIA systems. NVIDIA today released Nemotron 3.5 Super VL, an open 120-billion-parameter model that delivers the fastest
[7]
NVIDIA's Nemotron 3.5 Super VL with post-training beats Claude Opus 5, GPT-6.1 Sol on specialized task benchmarks
An enterprise agent reviewing a credit agreement may need to extract figures from tables, interpret a chart, check a related webpage and apply the company's own rules before it can produce a useful answer. NVIDIA's latest Nemotron includes three models that target this set of tasks and more: a
[8]
NVIDIA Expands Nemotron AI Models for Data Centres and Desktop Devices
NVIDIA announced three additions to its Nemotron family of open AI models. The models target enterprise workflows, local AI agents and conversational applications across data centres, desktop systems and edge devices. The launch includes Nemotron 3.5 Super VL, a 120-billion-parameter model, and
[9]
Nvidia expands Nemotron LLM family with hyper-efficient models for agentic systems
Nvidia expands Nemotron LLM family with hyper-efficient models for agentic systems Nvidia Corp. today announced the launch of customizable Nemotron 3.5 Super LV large language models for always-on artificial intelligence agents and edge-focused 3.5 Bolt for local systems. The company said the
[10]
Nvidia unveils 120B-parameter Nemotron Super VL
The 120-billion-parameter Super VL model targets enterprise workflows including document analysis, financial accounting, scientific research and security testing. Nvidia on Thursday launched open-weight artificial intelligence models for enterprise agents and local computing, including the
[11]
NVIDIA Announces The 120B Nemotron 3.5 Super VL Model For AI Agents And 2 Tiny 4 GB And 24 GB Edge Models Under The Nemotron 3.5 Bolt Umbrella
NVIDIA appears bent on supercharging its push towards open-weight models, especially after it recently inked an agreement to acquire Hugging Face, the platform that serves as a huge repository for just such AI models. As a case in point, consider that NVIDIA has just unveiled three new open-weight
[12]
NVIDIA Launches Nemotron 3.5 Super VL to Power Specialized AI Agents with Multimodal Reasoning
NVIDIA Nemotron 3.5 Super VL is an open, customizable multimodal model for these specialized workflows. Always-on agents are taking on longer, more complex enterprise workflows. A single request can involve reading a document, checking information across sources, following company policy, and
[13]
Nvidia Launches Open-weight AI Models to Run Agents on PCs within Enterprises
This creates a new market by selling personal compute powered by its own AI-enabled microprocessors that could run models locally and manage agents within its ecosystem Once upon a time there was a chipmaker who invested in the AI ecosystem in order to sell top-end processors so that their large
[14]
NVIDIA Nemotron 3.5 Super VL: Frontier level accuracy at one fifth the size
For many years in AI, the answer to everything has been bigger. Larger numbers of parameters, more GPUs, and larger bills. NVIDIA is saying the exact opposite. Its newest offering, the Nemotron 3.5 Super VL, has 120 billion parameters but claims it performs on par with its largest reasoning model
Share
Copy Link
Nvidia released three open AI models designed for agentic workloads: the 120-billion-parameter Nemotron 3.5 Super VL for enterprise agents and two lightweight Bolt models for local deployment on AI PCs. Super VL claims 5x faster task completion than competitors while maintaining accuracy approaching Nvidia's largest reasoning model. The release marks a strategic shift as US-based labs challenge China's dominance in open-weight AI development.
Nvidia Expands Open AI Models Portfolio with Speed-Focused Enterprise Solution
Nvidia has launched three new open AI models targeting different deployment scenarios, headlined by the Nemotron 3.5 Super VL, a 120-billion-parameter model built specifically for enterprise AI agents
1
. The company claims Super VL completes tasks five times faster than the nearest competitor in its class while delivering accuracy comparable to Nvidia's largest reasoning model despite using roughly one-fifth as many parameters5
. This positions the model as a practical choice for organizations seeking to deploy customizable AI models without sacrificing performance for efficiency.The release includes two additional variants under the Nemotron 3.5 Bolt family: the 74-billion-parameter Bolt 74B-A9B 24GB designed for local agentic tasks, and the compact Bolt E4B 4GB aimed at conversational agents
1
. These smaller models enable local deployment on AI PCs running Nvidia RTX Spark chips and workstations equipped with DGX Spark, allowing users to delegate everyday projects without sending workloads to the cloud1
. Super VL and Bolt 74B-A9B are available immediately, with Bolt E4B scheduled for release on October 231
.Enterprise Adoption Signals Confidence in Open-Weight AI Models
Major enterprises have already begun customizing Nemotron 3.5 Super VL for specialized applications across industries. Scale AI adapted the model for Skydance's vendor credit finance agent and EY's merger-and-acquisition due diligence agent
5
. Other early adopters include Ramp, Crowdstrike, Perplexity, and Salesforce, demonstrating confidence in the model's capabilities for document intelligence, financial accounting, scientific experimentation, medical insurance claims, and security attack resilience1
.AfterQuery customized Super VL for Hebbia's credit agreement analysis, while AI21 and Alice focused on cybersecurity applications
5
. Mercor has post-trained the model for coding tasks covered by the APEX-SWE benchmark5
. This diverse adoption pattern underscores the appeal of open-weight AI models that organizations can post-train on their own data, deploy on their own infrastructure, and own what they build—unlike closed frontier models1
.US Labs Challenge China's Open-Weight Dominance
The Nemotron 3.5 release represents another data point in the growing competition between US-based and Chinese AI labs in the open-weight space. China-based labs have historically held the upper hand on competitive open AI models, but recent Western releases are closing the geography gap
2
. French lab Mistral recently launched its Le Chonk model, achieving certain state-of-the-art benchmarks, while Reflection AI—in which Nvidia has invested and is in talks to acquire—released its Beam model last week2
.Artificial Analysis' benchmark testing positioned Super VL significantly ahead of competitors InclusionAI's Ling 3.0 Flash and DeepSeek's V4.1 Flash
2
. The industry is embracing open-weight AI models more readily following the OpenAI Hugging Face breach, which dispelled the notion that only open models posed security risks2
. If US-made open models prove themselves against Chinese alternatives, this could accelerate adoption in areas where cybersecurity guidance discourages using foreign technology2
.Expanded Model Portfolio Targets European Sovereign Intelligence
At GTC Berlin, Nvidia unveiled additional open-weight models spanning embedding, autonomous vehicles, and specialized applications to capitalize on European sovereign AI demand
3
. The refreshed Nemotron Retriever models support embedding and reranking for RAG workflows, handling both images and text across 34 languages3
. These models range from one to eight billion parameters, with NVFP4-quantized versions optimized for Blackwell-based accelerators3
.Nvidia claims its eight-billion-parameter embedding and reranking models now rank as the top dense models on the ViDoRe V3.1 benchmarks at BF16-precision
3
. The company also launched a refreshed Nemotron Parse model trained specifically to extract information from PDF documents, processing approximately 10.5 pages per second when running the one-billion-parameter model on a single H1003
. The Alpamayo 2 Edge model for autonomous vehicles promises real-time processing of streaming video and sensor data on Nvidia's DRIVE AGX Thor platform3
.Related Stories
Strategic Positioning Against Proprietary Frontier Labs
Nvidia's aggressive expansion into open AI models reflects a strategic response to shifting market dynamics. While the GPU giant has an entire family of Nemotron models covering multimodal artificial intelligence, reasoning, coding, math, and enterprise automation, proprietary AI labs—which are both Nvidia's customers and investments—are increasingly looking beyond Nvidia to power their needs
1
. AWS's Trainium, Cerebras, and home-grown chips are emerging as alternatives1
.Nvidia is leaning heavily into the trend of enterprise companies wanting to run their own open models in their own data centers for on-premise AI deployment
1
. The company's approach differs from most open-weight releases by including not just the model weights but also training data and recipes, allowing customers to build fully custom models based on Nvidia's Nemotron architecture rather than relying solely on post-training fine-tuning3
. Super VL's training process covered approximately 26 trillion tokens, with post-training involving around 3.8 million samples followed by reinforcement learning across dozens of environments5
.Local AI PC Market Addresses Privacy Concerns
With the smaller Bolt models, Nvidia is pushing forward the nascent trend of consumer autonomous agents while addressing privacy concerns. Privacy-conscious consumers may hesitate to invite AI agents deeply into their lives with access to family, health, and financial information
1
. Local deployment on AI PCs powered by RTX Spark chips offers an alternative that keeps sensitive data on-device1
.The Bolt 74B-A9B uses native 2-bit quantization to fit within 24GB of memory, allowing compatible systems to run the model alongside other applications for multistep reasoning, tool calling, and web browsing
5
. Nvidia ranks this model first among open models in its class on PinchBench and second place among open models on BrowseComp5
. The compact Bolt E4B targets conversational agents and physical AI systems running on Nvidia Jetson hardware5
. These AI PCs are being released through partners like Microsoft and Dell, aimed at consumers, programmers, and enterprise knowledge workers1
.References
Summarized by
Navi
[3]
[5]
Related Stories
Nvidia launches Nemotron 3 open source AI models as Meta steps back from transparency
15 Dec 2025•Technology
Nvidia launches Nemotron 3 Super with 5x throughput boost to power enterprise AI agents
11 Mar 2026•Technology
Nvidia releases Nemotron 3.5 Lightning and NeMo Switchyard to slash agentic AI costs by 66%
11 Aug 2026•Technology
Recent Highlights
Recent Highlights
Today's Top Stories
Don’t drown in AI news. We cut through the noise - filtering, ranking and summarizing the most important AI news, breakthroughs and research daily. Follow topics that matter to you and stay ahead.
