NVIDIA Nemotron 3.5 Super VL — 120B open multimodal model for agents
NVIDIA released Nemotron 3.5 Super VL, a 120B-total/~12B-active open-weight mixture-of-experts model for multimodal agentic enterprise workloads spanning text, images, documents, and video, alongside two Bolt edge models: 74B-A9B for 24GB and E4B for 4GB footprints.
Why it matters
OpenMDW-licensed 120B model NVIDIA claims beats Claude Opus 5 and GPT-6.1 Sol on specialized agentic benchmarks, with 5x the throughput of peers in its class; built with the Nemotron Coalition (Mistral AI, Thinking Machines Lab, Perplexity, Black Forest Labs).
What you could build with it
A small B2B SaaS team selling document automation to finance firms could fine-tune Super VL on client report formats and run it on their own GPUs, replacing a patchwork of OCR plus vision-API calls with one model that reads charts, tables, and scanned pages in a single pass.
Does it hold up?
Too early to judge: released today. NVIDIA lists dozens of companies already fine-tuning it (Scale AI, AfterQuery, TrainLoop), but independent verification of the vendor's frontier-beating benchmark claims is still pending.
Built with NVIDIA Nemotron 3.5 Super VL
- NVIDIA Announces The 120B Nemotron 3.5 Super VL Model For AI Agents And 2 Tiny Edge Modelsarticle · Hardware-angle coverage: trained on 25T tokens plus 4M post-training samples, packaged as a production-ready NIM microservice, running across DGX Spark, DGX Station, Jetson, and RTX hardware.
- NVIDIA launches open Nemotron 3.5 Super VL and Bolt edge modelsarticle · Notes early traction: NVIDIA lists dozens of companies fine-tuning Super VL on private data, including Scale AI for finance agents, plus Atlassian, CrowdStrike, Salesforce, and Perplexity integrating it into agent products.
Learn more
First spotted on article: source.
More AI models
OpenAI publishes 722 AI-generated math manuscripts from an unreleased modelOpenAI released 722 mathematical manuscripts in 372 result families to GitHub under Apache-2.0, produced by an…models · JEV 0.74StepFun Step-5-Preview — open BF16 weights releasedStepFun published the full BF16 weights for its flagship Step-5-Preview agentic model (600B-parameter sparse MoE, 27B…models · JEV 0.73ElevenLabs Eleven v4 + v4 Turbo: new emotive TTS architecture with a 100ms real-time variantElevenLabs released Eleven v4 and Eleven v4 Turbo on September 28: a new text-to-speech architecture with inline…models · JEV 0.72Mistral Large 4 (Le Chonk): 1.05T-param open-weight MoEMistral's new flagship: a 1.05T-parameter mixture-of-experts multimodal model with 49B active params and 1M-token…models · JEV 0.72VoiceStudio: open-source, fully-local ElevenLabs alternative with voice cloning, dubbing and dictation in 646 languagesVoiceStudio is an open-source, fully-local voice AI studio: voice cloning from a 3-second sample, voice design, video…models · JEV 0.71JetBrains Mellum 2.1: 12B MoE coding model trained for agentsJetBrains released Mellum 2.1, a 12B MoE model (2.5B active parameters, 128K context) under Apache 2.0, trained mainly…models · JEV 0.69
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.