Nvidia releases open Nemotron 3.5 Super VL and two Bolt models
For the CTO of a regional bank
Nvidia said in a blog post dated Oct. 15 that it released Nemotron 3.5 Super VL, an open 120-billion-parameter model, along with two edge models, Nemotron 3.5 Bolt 74B-A9B 24GB and Bolt E4B 4GB.1
Nvidia said organizations can post-train Super VL on their own data and deploy it on their own infrastructure, unlike closed frontier models.1 The model ships with open weights, open datasets and training recipes, according to Nvidia; it did not state a commercial motive for the release.1
Nvidia said Super VL is comparable on multimodal agentic tasks with its largest reasoning model at roughly one-fifth the parameter count, and up to 5x faster in throughput than other models in its class on the Artificial Analysis Intelligence Index leaderboard.1
Nvidia named Ramp as customizing the model for financial document intelligence, and said AfterQuery customized it for Hebbia's credit agreement analysis; Scale AI post-trained it for Skydance's vendor credit finance agent and EY's M&A due diligence agent.1
Nvidia said the model reads OCR text, tables, charts, webpages and PDFs, and is available now on Hugging Face, Open Router, the ModelScope community and as NVIDIA NIM, which carries enterprise-grade support and security patches every two weeks.1
A model listing quoted on an Nvidia developer forum post says the model has 120B total and 12B active parameters and supports context lengths up to 1M tokens; the post is a user's, not an official announcement.3
A bank can run this model inside its own environment and keep customer records on site, which suits data-residency and privacy requirements. It also shifts vendor risk, GPU capacity and model validation onto the bank's own teams. The sources give no figures for hardware needs or cost.
Bolt E4B 4GB is due Oct. 23, and Nvidia CEO Jensen Huang's GTC Berlin keynote is set for Oct. 21.1
Our take. Closed models ask you to send the loan file to the model. Super VL lets you bring the model to the loan file, and then you own the servers, the tuning and the validation.
Action. Model risk: Scope validation and data-lineage requirements for an on-premises, bank-fine-tuned open-weight model before any credit-review pilot.
What our records say
Ramp is a business finance and infrastructure fintech, and the records describe its AI as agents across the finance back office, plus products that manage AI spending itself.1 Ramp sells agents for expense review, bills, procurement and accounting, and since April 2026 tools to track and route companies' AI token spend.1 The records do not describe a bank-specific AI record. Use cases and statuses: Policy agents for expense review: Rolling out.2
From our records
- Rampfintech · company Business finance & infrastructure, founded 2019, United States
Sources
- Primary sourceOpen From Data Center to Desktop: NVIDIA Expands Nemotron Family With Highest Efficiency Super VL and Bolt Models | NVIDIA Blog Nvidia
- Primary sourceWhat are Open Models? | NVIDIA Glossary Nvidia
- Primary sourceNVIDIA Nemotron-3.5-Bolt-74B-A9B-24GB and Nemotron-3.5-Super-VL-120B-A12B published - NVIDIAs new general purpose models especially for GB10 Nvidia