AnswersPublished15 sources
Nvidia’s Nemotron 3.5 Models: What’s Available and What’s Coming
Nvidia announced three open Nemotron 3.5 models on Oct. 15, 2026: Super VL for specialized agents, Bolt 74B A9B sized for 24GB memory and a 4GB Bolt E4B.
Published byEdited with GPT-6 LunaImages generated with GPT Image 2
Research answer
Key takeaways
- Nvidia announced three open Nemotron 3.5 models on Oct. 15, 2026: Super VL for specialized agents, Bolt 74B A9B sized for 24GB memory and a 4GB Bolt E4B.
- The main distinction is deployment: Super VL targets specialized, always on agent workloads, while the Bolt models are designed for local or edge use.
- Nvidia reports partner post training improvements for specialized workflows, but those figures should be treated as company reported results—not universal performance guarantees.
AI Prompt
openai.comCreate a landscape editorial hero image for this Studio Global article: What did Nvidia announce on Oct. 15, 2026, about its three open-weight Nemotron 3.5 models for AI agents spanning data centers and local har. Article summary: On Oct. 15, 2026, Nvidia announced three open Nemotron 3.5 models intended to extend AI agents from data centers to desktops and edge devices: Super VL, Bolt 74B-A9B 24GB and Bolt E4B 4GB. Super VL and Bolt 74B-A9B were . Topic tags: general, general web, documentation, user generated. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks,
Nvidia’s Nemotron 3.5 announcement is built around a practical split: a large model for specialized AI agents and two smaller models aimed at running agents closer to the user. Super VL and Bolt 74B-A9B were available on Oct. 15, 2026; the smaller Bolt E4B was scheduled for Oct. 23. 10
Three models, two deployment scales
| Model | Intended role | Size or memory target | Availability at announcement |
|---|---|---|---|
| Nemotron 3.5 Super VL | Specialized, always-on agents | 120 billion parameters | Available Oct. 15 |
| Nemotron 3.5 Bolt 74B-A9B | Local agent workloads | 74 billion total parameters, 9 billion active; designed to fit in 24GB | Available Oct. 15 |
| Nemotron 3.5 Bolt E4B | Fast conversational agents on local or edge hardware | 4GB footprint | Scheduled for Oct. 23 |
Nvidia describes Super VL as an open model for specialized, always-on agents. Its announcement positions the model alongside Bolt 74B-A9B, which is optimized for local agent workloads, and Bolt E4B, which is designed for fast conversation on a smaller memory budget. 10
The Bolt 74B-A9B model has 74 billion total parameters, with 9 billion active per token. Nvidia says it uses two-bit quantization to fit within 24GB of memory and is designed for local agent tasks on RTX and DGX Spark systems. 6 The 4GB Bolt E4B is the smallest of the three, and was announced as a later release rather than an Oct. 15 launch.
10
What Nvidia says about customization and performance
Nvidia says organizations including Scale AI, Ramp and AI21 have post-trained Super VL for specialized workflows. Its developer material reports improvements that include 40-fold lower inference cost, tenfold lower latency and 40% higher accuracy. These are company-reported results from specialized post-training; the cited material does not establish that every organization achieved every figure or that the gains apply to other deployments. 17
Nvidia also presents Super VL as a fast model for agent workloads. ZDNET reported the company’s claim that it offers five times the throughput of the next-best model in its class. That comparison is Nvidia’s claim, not an independent guarantee across hardware, workloads or evaluation methods. 9
Why the release matters
The announcement extends Nvidia’s open-model lineup across server-scale and local deployments. Super VL is aimed at specialized agent work, while Bolt 74B-A9B and Bolt E4B are positioned for local and edge systems with tighter memory constraints. 3
10
That gives developers more than one deployment target within the same model family. It does not, by itself, show that local AI will reduce or increase demand for Nvidia chips: the announcement establishes the range of intended deployments, not their eventual commercial impact. 6
10
Studio Global AI
Continue your research
This page includes a source-backed answer you can continue inside Studio Global.
People also ask
What is the short answer to "Nvidia’s Nemotron 3.5 Models: What’s Available and What’s Coming"?
Nvidia announced three open Nemotron 3.5 models on Oct. 15, 2026: Super VL for specialized agents, Bolt 74B A9B sized for 24GB memory and a 4GB Bolt E4B.
What are the key points to validate first?
Nvidia announced three open Nemotron 3.5 models on Oct. 15, 2026: Super VL for specialized agents, Bolt 74B A9B sized for 24GB memory and a 4GB Bolt E4B. The main distinction is deployment: Super VL targets specialized, always on agent workloads, while the Bolt models are designed for local or edge use.
What should I do next in practice?
Nvidia reports partner post training improvements for specialized workflows, but those figures should be treated as company reported results—not universal performance guarantees.
Sources
blogs.nvidia.comOpen From Data Center to Desktop: NVIDIA Expands Nemotron Family With Highest Efficiency Super VL and Bolt Models
research.nvidia.comNVIDIA Nemotron 3.5 Bolt 74B-A9B 24GB - NVIDIA Nemotron
developer.nvidia.comNVIDIA Nemotron 3.5 Super VL Powers Specialized ...
zdnet.comNvidia’s new Nemotron 3.5 Super VL is another step forward for US open-weight AI