Cloudflare runs open models on its own network under the name Workers AI, with a free daily allowance. It is a little different from other providers: an app needs two values from you, an API token and your Account ID. Most setup problems come from entering only the first.
Create the API token
- Sign in to the Cloudflare dashboard and open the Workers AI page.
- Select Use REST API.
- Select Create a Workers AI API Token and review the prefilled details.
- Select Create API Token.
- Select Copy API Token and store the value somewhere private.
The prefilled template is the easy route. If you create a custom token instead, it needs both the Workers AI - Read and Workers AI - Edit permissions.
Find your Account ID
On the same Workers AI page, the Get Account ID step shows your Account ID: a string of 32 letters and digits. Copy it as well. It is not secret in the way the token is, but an app cannot reach Workers AI without it.
What the free allowance means
Cloudflare measures usage in Neurons, a unit for the computing work a request needs, not in requests.
- You get 10,000 Neurons per day at no charge.
- The allowance resets at 00:00 UTC.
- On the Workers Free plan, requests fail once the allowance is used up, until the reset.
- On Workers Paid, usage beyond the allowance costs $0.011 per 1,000 Neurons.
How many questions that is depends on the model and on the length of the question and answer. Cloudflare publishes a Neuron price for each model:
| Model | Neurons per million input tokens | Neurons per million output tokens |
|---|---|---|
| Gemma 4 26B | 9,091 | 27,273 |
| GLM 4.7 Flash | 5,500 | 36,400 |
As a rough example, a text question of about 1,000 tokens with a 500-token answer costs about 23 Neurons on Gemma 4 26B, so the free 10,000 cover a few hundred such questions a day. Image questions send more input tokens and use more. The dashboard shows what you have used today.
Which model to pick
- Gemma 4 26B accepts images and handles text well. A good default.
- GLM 4.7 Flash is a fast text model.
Using the token in LuciQA
LuciQA is a desktop app that answers questions about what is on your screen. Open Settings, pick Cloudflare, paste the API token and the Account ID, choose a model and save. If you choose GLM and then send an image, LuciQA switches to Gemma for that request, because GLM reads text only.
If saving fails with a message about the account, check that the Account ID is exactly 32 characters with no spaces. The Cloudflare setup guide repeats these steps, and the free AI APIs comparison shows how the other free providers count usage.