Ollama

ApplicationOllama

Also called ollama windows, ollama.exe

OllamaProductivity

A local model runner. The Windows process is ollama.exe. Model files are large. It is not the ChatGPT desktop app.

What you need to know

Ollama runs models on this PC. ChatGPT.exe is a cloud client.

ollama.exe is a local runner. Model files are large. ChatGPT desktop talks to OpenAI. Claude desktop talks to Anthropic. LM Studio is a different local GUI for GGUF files.

What is it?: Local model runner
Same as ChatGPT?: No. This runs on your PC
Are models small?: No. Files are large

Feex says

If ollama.exe is using RAM, I would check which model is loaded. Those weights live on disk and in memory. I would not end ChatGPT.exe to free that. Different product, different vendor.

What is Ollama?

Ollama runs local models through ollama.exe. It is not ChatGPT desktop, not Claude desktop, and not LM Studio's GUI.

If you have never used it

If you want a model to run on this PC, Ollama is a common runner. Download it from ollama.com/download. ollama.exe is a service plus CLI. The model files are large and live under your user profile. ChatGPT.exe talks to OpenAI in the cloud. A browser tab on chatgpt.com is not this exe.

What is it used for?

Pull and run a model

ollama run llama3.2. ChatGPT.exe will not download that file into .ollama.

Local API on localhost

Apps talk to Ollama's port. OpenAI's cloud API is a different endpoint.

Keep the service in the tray

ollama.exe stays so models can load. LM Studio is a separate GUI product.

Who is it for?

People who want a local model without a ChatGPT account
Developers wiring an app to localhost
Anyone who thought ChatGPT desktop was a local model

Current status

Ollama (2023) is a current local model runner from Ollama. The vendor still points Windows users at https://ollama.com/download.

How old is it?

Ollama launched in 2023 as a local-model helper. It is the same generation as LM Studio and newer than ChatGPT the web product. Current Windows builds still update. A 2022 llama.cpp-only zip is a different packaging.

Versions

Major channels first. Exact build numbers are not all verified here.

Ollama for Windows

System requirements

Values follow the publisher's public guidance. Workload still changes the real need.

Minimum

Minimum system requirements
Operating system	A current 64-bit Windows 10 or Windows 11 release, per Ollama
RAM	The runner is small. The model you load can want several to tens of GB

Resource profile

CPU demand: Follows what you open inside the app
RAM demand: Follows documents, tabs, or libraries

Installed processes

These executables belong to this product when it is installed.

Installed processes
ollama.exe	Ollama service and CLI

Startup components

Ollama can start a service or tray helper at sign-in so the API is ready. That is not ChatGPT's startup toggle.

File locations

Common Ollama locations
Typical install	C:\Users\<you>\AppData\Local\Programs\Ollama\

File formats

GGUF and Ollama model blobs

Common problems

Disk shock after ollama pull

Models are multi-gigabyte. That is expected. ChatGPT desktop does not store those weights.

GPU not used

Ollama documents GPU backends. A CPU-only run is slow. That is this runner, not an Edge tab.

Getting unstuck

Install from ollama.com/download

Ollama hosts the Windows installer. A random ollama.exe in Downloads is the case to inspect.

ollama list and the .ollama\models folder

See what you actually pulled. LM Studio keeps models in a different cache.

Quit the tray app if you want the service gone

Ending ChatGPT.exe will not stop ollama.exe.

Alternatives

LM Studio if you want a GGUF GUI. ChatGPT or Claude if you want a cloud model. llama.cpp by itself is a lower-level path.

Official help and downloads

Start with the vendor. Random exe sites are not a support channel.

Official contacts
Download	ollama.com/download Official installer or vendor page
Help	github.com/ollama/ollama/tree/main/docs Vendor support
Official site	ollama.com/ Start here, not a random exe host
Ollama GitHub docs	github.com/ollama/ollama/tree/main/docs CLI, models, and GPU notes

Related hardware

RAM and disk decide which models fit. A supported GPU helps. This is local compute, not a browser chat.

Technical details

Vendor version numbers change. This article does not pin a build.

Sources and verification

Reviewed by Feex editorial. Last verified August 2026 Status: Feex Reviewed. How we research.

Ollama official site · Vendor product page
Ollama help · Vendor support
Ollama documentation · Official project docs
Foxhole process map · First-party samples on Windows PCs. Not a lab benchmark.

Was this article useful?

A yes or no is enough. Pick a reason if you want, or add a short note.