How to Rename Files with Ollama on Mac (Offline AI Guide)

Quick answer: rename files locally on Mac with Ollama by installing Ollama, pulling a vision-capable model such as qwen2.5vl:3b, gemma3:4b, or granite3.2-vision:2b, then enabling Offline AI mode in Zush. Supported analysis content stays on your device and there are no provider API costs.

Mac users finally have a practical way to rename files with AI without routing supported analysis content through a cloud provider. Zush Offline AI mode connects to Ollama, which runs a local model on your Mac and returns filename suggestions for screenshots, videos, images, PDFs, and supported documents.

This guide is the hands-on Mac setup tutorial for Ollama: install, pull a model, connect, test. If you are still deciding whether to run models locally at all, read when local AI file renaming is worth the tradeoff first and come back here for the commands. Zush for Windows is live, but this article does not assume the Windows Offline AI setup flow is identical. For the current app page, see Zush for Mac. For the broader product surface, see File Renamer for Mac and Windows. For the short setup reference, open the Ollama setup guide. If you are comparing local-first alternatives, read Zush vs AI Renamer.

Looking to rename an Ollama model instead?

If you arrived here searching for how to rename a model in Ollama, that is a different task: Ollama has no dedicated rename command. The supported approach is to copy the model under the name you want, then remove the original.

ollama cp llama3 my-custom-name
ollama rm llama3

If you are here to rename files using a local Ollama model, keep reading.

What you need before starting

You need three pieces:

• Zush for Mac with Offline AI mode available on macOS 15 Sequoia or later
• Ollama for macOS
• a vision-capable Ollama model

Ollama has its own macOS requirement on its download page. Zush for Mac requires macOS 15 Sequoia or later and is built for modern macOS workflows with Finder, Spotlight-friendly metadata, batch rename, folder monitoring, and one-click rename history.

The important part is the model. Text-only models are not enough for file renaming because many files are visual: screenshots, videos, photos, scans, and document previews. Start with a small vision model before trying anything large.

Step 1: install Ollama

Download Ollama from the official Ollama download page, install it, and open it once. Opening the app starts the local runtime that Zush talks to.

If you prefer Terminal, Ollama also documents command-line installation on its download page. For most Mac users, the regular macOS app install is the simpler path.

Step 2: pull a vision model

Open Terminal and download one model. For most users, start with:

ollama pull qwen2.5vl:3b

The qwen2.5vl page lists qwen2.5vl:3b as a text-and-image model, which is why it is a practical first choice for screenshots and images.

Other good options:

ollama pull gemma3:4b
ollama pull granite3.2-vision:2b

gemma3 includes multimodal vision variants such as gemma3:4b. granite3.2-vision is a compact vision-language model designed around visual document understanding.

Start with one model, not five. Once the first one works, you can compare speed and naming quality.

Step 3: confirm Ollama is running

In Terminal, run:

ollama list

You should see the model you pulled. If the list is empty, the model download did not finish or the model name was mistyped.

If Zush cannot connect, make sure Ollama is open. You can also start the local service manually:

ollama serve

The default local host used by the setup guide is http://127.0.0.1:11434.

Step 4: enable Offline AI mode in Zush

Open Zush and go to BYOK/Offline. Refresh the model list, choose the model you installed, run the connection test, and then switch on Offline AI mode.

Offline AI mode is separate from Zush Cloud and BYOK:

If you process a lot of files and still want cloud models, read BYOK: Unlimited AI File Renames with Your Own API Key. If you are not yet sure local processing is the right mode for your files, step back to when local AI file renaming is worth the tradeoff, which covers what local models handle well and where cloud AI is still stronger, or the broader Cloud AI vs Local AI File Renaming.

Step 5: test with a small batch

Do not test on a huge work archive first. Pick a sample folder with 10 files:

• a few screenshots
• one or two photos
• a PDF or scan
• a document export

Run the rename preview and inspect the proposed names. Good names should be descriptive without being wordy:

If the names are too long, tighten your custom prompt. If they are too generic, try another model. For naming conventions, use File Naming Conventions: Best Practices for Searchable Files.

Best Mac folders to start with

Screenshots

The macOS screenshot folder fills with timestamp names. Local AI can turn those into names based on visible UI, charts, messages, and documents. For a deeper workflow, see How to Rename Screenshots Automatically on Mac.

Downloads

Downloads is where random filenames collect. Local AI helps when files are visual or text-readable. Pair it with a weekly cleanup routine from How to Organize Your Downloads Folder on Mac.

Scans and PDFs

PDFs often arrive as scan_001.pdf or document.pdf. A local vision model can help when the document preview contains enough visible context. For PDF-specific guidance, read Rename PDF Files with AI on Mac.

Photo imports

For casual photo folders, local AI can replace IMG_ names with searchable descriptions. For large professional photo archives, test carefully because local model speed can vary by hardware and image size. For more photo workflow detail, read Best Ways to Organize Photos on Mac.

Troubleshooting

Zush does not see any models

Run ollama list. If no models appear, pull a model first. Then refresh the model list in Zush BYOK/Offline.

Connection test fails

Make sure Ollama is open. If needed, run ollama serve and confirm the host setting in Zush points to http://127.0.0.1:11434.

Renaming is slow

Try a smaller model, close memory-heavy apps, or switch back to Cloud for large batches where speed matters more than local processing.

Names are vague

Try a stronger model, add a custom prompt, or use Cloud/BYOK for files where naming quality matters more than offline processing.

FAQ

Can I rename files with Ollama on any Mac?

You need a Mac that can run the current Ollama macOS app and has enough memory and disk space for the model you choose. Smaller models are a better starting point for everyday Macs.

Does Zush send my files to the cloud in Offline AI mode?

For supported Offline AI processing, analysis content is processed by the local Ollama model instead of Zush cloud or third-party AI providers. Licensing, updates, support, or non-content operational checks may still use backend services. See the privacy policy for details.

Should I use Ollama or BYOK?

Use Ollama when local processing is the priority. Use BYOK when you want unlimited cloud renames with provider-level billing and model quality. Both are Pro-oriented workflows in Zush.