Contents

Ollama Cloud Models Tutorial: Run Large Models with the Same Commands

The models you can run locally are limited by your GPU memory. Large models with hundreds of billions of parameters simply won't fit on a typical computer. Ollama cloud models let you run these large models on Ollama's servers with the same ollama run command and the same API.

Local models vs. cloud models

	Local models	Cloud models
Where they run	Your computer	Ollama's servers
Download required	Yes, possibly tens of GB	No
Model size	Limited by your memory	You can use very large models
Internet required	No	Yes
Account required	No	You need an ollama.com account
Data	Stays on your computer	Sent to Ollama's servers for processing
Cost	Free	Usage limits apply, depending on the plan

How to identify cloud models

On model pages at ollama.com, models that support cloud execution are marked cloud, and their tag names also include cloud. You can filter by the Cloud category in the model search on ollama.com to see the cloud models currently available.

Get started

ollama signin

Follow the prompts to sign in to your ollama.com account in your browser. Newer versions of Ollama also guide you through signing in the first time you run it.

2. Run a model

ollama run model-name:tag-with-cloud

You use cloud models exactly like local models. You can also use a cloud model's name directly through the local API (http://localhost:11434), and Ollama automatically forwards the request to the cloud. This means Open WebUI or your own program can use cloud models without any changes.

Call the cloud API directly

You can call the ollama.com API directly without installing Ollama:

Create an API key in your ollama.com account settings.
Set an environment variable:
```
export OLLAMA_API_KEY="your-api-key"
```

Call the API:

curl https://ollama.com/api/chat \
  -H "Authorization: Bearer $OLLAMA_API_KEY" \
  -d '{
    "model": "cloud-model-name",
    "messages": [{ "role": "user", "content": "Hello" }],
    "stream": false
  }'

Here's how to use the Python package:

import os
from ollama import Client

client = Client(
    host="https://ollama.com",
    headers={"Authorization": "Bearer " + os.environ["OLLAMA_API_KEY"]},
)
response = client.chat(model="cloud-model-name", messages=[{"role": "user", "content": "Hello"}])
print(response.message.content)

The OpenAI-compatible endpoint is https://ollama.com/v1.

Don't put your API key in frontend code or commit it to Git.

List the cloud models currently available:

curl https://ollama.com/api/tags

Usage and plans

Cloud models have usage limits. You can check your current usage in your ollama.com account settings. See the official pricing page for details about plans. Cloud models may also be retired over time; your account settings show models that are about to be retired. Models already downloaded to your computer are unaffected.

Privacy considerations

When you use cloud models, your prompts and responses are sent to Ollama's servers for processing. Ollama says it does not use this data to train models, but if you are handling confidential data (such as customers' personal information or unpublished business information), use a local model or check your company's data policy first.

Disable cloud features completely

In an environment that allows local computation only, you can disable all cloud features (cloud models and web search):

OLLAMA_NO_CLOUD=1

See Environment variables and settings for how to configure this.

When to use cloud models

Local model quality isn't good enough, and you need a larger model
You're using a laptop or a computer without a dedicated GPU
You want to try a large model before deciding whether to invest in hardware

For everyday, high-volume, or confidential tasks, local models are still recommended.

Updated October 12, 2026