Ollama 0.40.10 adds Nemotron 3.5 Bolt E4B, the small Bolt that also runs on Apple Silicon

Ollama 0.40.10 adds Nemotron 3.5 Bolt E4B, NVIDIA's smaller, Gemma 4-based vision-language model for fast local conversations, image understanding and tool use. Unlike the big Bolt that came with Ollama 0.40.5, this one is supported on Apple Silicon as well as on Windows and Linux , and its download is less than a third the size. The release page carries the date 23 October, 15:03 ; the installers were uploaded the same day at 17:09 UTC , and the page's changelog line runs from v0.40.9 to v0.40.10.

What changed

One model, one command. The release page lists one addition, and its command is ollama run nemotron-3.5-bolt:e4b. Vision-language means the model reads images as well as text: you can hand it a screenshot or a photo and ask about it. It supports a 128K-token context window . A token is a word or a piece of one, and the context window is how much text the model can hold at once, the conversation so far included.

Ollama's library page adds the detail. Bolt E4B is an 8-billion-parameter model, with 4 billion effective, post-trained by NVIDIA from Gemma 4 E4B. Parameters are the numbers a model learned. Post-trained means NVIDIA started from an existing model, Gemma 4 E4B, and trained it further. It is designed for fast local conversations, instruction following and multi-turn tool use : tasks in which the model calls a function, reads the result and carries on over several turns.

A draft that speeds it up.The model includes a DFlash draft for speculative decoding. A small draft model guesses the next few words and Bolt E4B checks the guess, keeping what it agrees with, so text comes out faster; the library page says the companion draft accelerates generation in exactly that way.

Size and systems. Ollama's library lists nemotron-3.5-bolt:e4bat 7.2GB - 7.9GB, with a 128K context window, for text and images. E4B is supported on both Windows/Linux and Apple Silicon. The models on the page are released under the OpenMDW License Agreement, version 1.1.

What it was before

The page's changelog line starts at v0.40.9 ; our item onOllama 0.40.9 covers that release. The Bolt that arrived with Ollama 0.40.5, covered in our item on Ollama 0.40.5, is the large one: 74 billion total parameters, with 9 billion active per token. It is still the library'snemotron-3.5-bolt:latest, listed at 25GB with a 256K context window, for text and images , and it is what you get if you type the name without:e4b. Apple Silicon and Metal, the graphics layer Apple's chips use, are not currently supported for it. So on a Mac, E4B is the Bolt you can run.

What it means for you

If you have a Mac with Apple Silicon,this is your Bolt: E4B is supported on your machine , and the big one is not. Start withollama run nemotron-3.5-bolt:e4b, and keep the:e4bat the end; without it you pull the 25GB model that your Mac does not support.

Check your memory before you pull it.The download is 7.2GB - 7.9GB. A model runs well only when its file fits in the memory of the part doing the work — a PC's graphics card, or a Mac's unified memory, which the processor and graphics cores share — and the file is only the start. The library page says memory requirements depend on context size, visual inputs and speculative decoding. All three apply to E4B: its context window reaches 128K tokens , it reads images , and its draft runs beside it. If the file only just fits, begin with short, text-only conversations, and add images and long documents once you see how it runs. Our guideWhich Machine Runs a Model Locally explains that arithmetic on the machine you already have, and Mac or PC for Running Models Locally explains why the two machines differ.

If you are on Windows or Linux,E4B is supported on your system too. If the 25GB Bolt was more than your graphics card could hold , E4B is the size to try; in exchange for the smaller file you give up the 256K context window for 128K.

What to use it for.It is built for quick conversations, following instructions and tool use over several turns , and for questions about images.

How to update.The model arrives with this release , so update Ollama first. The release's list of files carries an installer for each system, all uploaded on 23 October. On Windows, download and runOllamaSetup.exe; on a Mac,Ollama.dmg; on Linux, theinstall.shscript from the same list. Then open a terminal and typeollama -v: it prints the version you have installed, which should now be 0.40.10. If it shows an older number, quit Ollama completely, start it again and run it once more.

The source

The release page on GitHub, Ollama v0.40.10, its list of downloadable files with their upload times, and the model's page in Ollama's library, nemotron-3.5-bolt.