NVIDIA releases open AI models, one built to run on your own PC
NVIDIA says its new Super VL model is five times faster than rivals, and its Bolt model can run on a home PC without the cloud.
NVIDIA has released two new kinds of AI models. Anyone can download them, run them on their own machines and change them. That is what "open" means here: NVIDIA shares the model itself.
The big one is Super VL. It reads images, web pages and PDFs, not just text. That includes tables and charts. NVIDIA says it is about as accurate as its own largest reasoning model, at roughly one-fifth the size. It also says Super VL is five times faster than the next best open model in its class.
The second is Bolt. It is built to run at home. The version out now is squeezed to fit in 24GB of memory. NVIDIA says it can research the web, write and fix code on your own machine, with no cloud. A smaller version for voice assistants is due Oct. 23.
NVIDIA says Atlassian, CrowdStrike, Perplexity and Salesforce are building on Super VL. It also says Scale AI tailored the model for work at EY and Skydance.
To be honest, most people won't notice a change today, and the speed claims are NVIDIA's own. Companies with private data can now run Super VL on their own servers. Hobbyists with 24GB of memory can try Bolt now, and their files never leave their machine.
The facts
- NVIDIA released Nemotron 3.5 Super VL, an open model with 120 billion parameters.
- NVIDIA says Super VL is comparable in accuracy to its largest reasoning model at roughly one-fifth the parameter count.
- NVIDIA says Super VL is 5x faster than the next best open model in its class.
- Nemotron 3.5 Bolt 74B-A9B 24GB uses 2-bit compression to fit in 24GB of memory. It is available now on Hugging Face.
- Bolt E4B 4GB is coming Oct. 23. It is aimed at local voice assistants, interactive characters and embedded interfaces.
- NVIDIA says Bolt 74B-A9B is the #1 open model in its class on PinchBench and the #2 open model on BrowseComp.
- Super VL is available on Hugging Face, Open Router, the ModelScope community and as NVIDIA NIM. NVIDIA says it ships with open weights, open datasets and training recipes.
- NVIDIA says Atlassian, CrowdStrike, Dassault Systèmes, Perplexity and Salesforce are integrating or customizing Super VL. Scale AI post-trained it for EY and Skydance.
Sources
- NVIDIA blog: Open From Data Center to Desktop: NVIDIA Expands Nemotron Family With Highest Efficiency Super VL and Bolt Models
Published
More in New models
AmazonNew modelsAyobi’s pickReflection AI's Beam, a free-to-download coding AI, is now available on Amazon's cloud
Reflection AI released Beam, an AI built to write and fix software over long jobs. Anyone can download and change it, and Amazon now offers it to run on its cloud.
New modelsOfficial releaseAyobi’s pickvia Hugging FaceIBM releases AI models that chat, make quick calls and transcribe speech in one
IBM says its new Granite Switch 4.1 models handle conversation, yes-or-no style judgments and speech-to-text from one online connection, using the memory of one model.
New modelsvia The DecoderBlack Forest Labs releases FLUX 3 Image Klein, an image AI for home computers
The new model makes and edits pictures up to 4K size. Black Forest Labs says it runs on an NVIDIA graphics card with 8 GB of memory.