Cloudflare Workers AI: 10,000 Free Neurons a Day, Then $0.011

As of October 2026, Cloudflare Workers AI gives every account 10,000 Neurons per day at no charge, and anything above that costs $0.011 per 1,000 Neurons on the Workers Paid plan. The free allocation resets daily at 00:00 UTC. If you only need the number to plan around: 10,000 free Neurons per day is 300,000 over a 30-day month, and the first 1,000,000 Neurons beyond that cost $11.

This post covers the pricing rules, the rate limits that bite before the money does, and the smallest Worker that calls a model.

What you get for free, and what happens at the limit

Workers Free accounts get the same 10,000 daily Neurons. Going past that requires upgrading to Workers Paid, which has a $5 per month minimum charge. Some models also accept prepaid AI Gateway credits as an alternative to the Paid plan.

A quick budget check, using only the prices above:

• 1,000,000 Neurons above the free allocation = 1,000 x $0.011 = $11.
• Because the allocation resets daily at 00:00 UTC, a batch job that straddles midnight UTC draws on two separate daily allocations.

Per-token prices differ a lot by model

The same Neuron billing sits behind very different per-token prices. Input and output are priced separately, per million tokens:

The cheapest and most expensive rows differ by about 49x on input price ($1.320 divided by $0.027). Pick the smallest model that passes your own test set before you look at anything else.

Rate limits hit before the bill does

Text generation is limited to 300 requests per minute, unless the model requires the Workers Paid plan. For models that do require it, the limit is per account and per model:

• 20 requests per minute on standard Workers AI billing.
• 50 requests per minute with prepaid AI Gateway credits (unified billing).

Other task types have their own ceilings, for example text embeddings at 3,000 requests per minute (1,500 for @cf/baai/bge-large-en-v1.5) and translation at 720 requests per minute. A queue consumer that fans out 100 text-generation calls in one minute to a paid-only model will be throttled at 20 or 50, whatever your budget says.

The smallest Worker that calls a model

Add the binding to your Wrangler config:

{
  "ai": {
    "binding": "AI"
  }
}

Then call a model from the fetch handler through env.AI.run. The model name below is the one used in Cloudflare's getting-started guide:

export default {
  async fetch(request: Request, env: { AI: Ai }) {
    const response = await env.AI.run("@cf/google/gemma-4-26b-a4b-it", {
      messages: [
        { role: "system", content: "Answer in one sentence." },
        { role: "user", content: "What is a Neuron in Workers AI billing?" },
      ],
    });
    return Response.json(response);
  },
};

Deploy with npx wrangler login and then npx wrangler deploy. npx wrangler dev runs locally, but local runs still use your Cloudflare account and incur usage charges, so the free allocation is spent while you test.

What I would do first

1. Run your real prompts on the smallest model in the table and compare answers before moving up.
2. Log the Neuron usage per call for a day so the $0.011 per 1,000 figure turns into your own number.
3. If a feature needs a paid-only model, plan around 20 requests per minute, not 300.

I run my own site, my-blog.org, on Cloudflare Workers with D1 and R2. The browser tools on it are served from that stack, and you can try them at my-blog.org/tools.

Limits of this post

Prices change. Every number here is from Cloudflare's documentation as of October 2026; check the Workers AI pricing page before you commit a budget. I did not convert Neurons to tokens, because the documentation page I used does not publish a conversion.

FAQ

Is Workers AI free? Yes up to 10,000 Neurons per day, on Workers Free or Paid. Beyond that you need Workers Paid at $0.011 per 1,000 Neurons.

When does the free allocation reset? Daily at 00:00 UTC.

Do I need the Paid plan to try it? No. The Free plan includes the 10,000 daily Neurons.

Why do I get throttled before I run out of Neurons? Per-minute rate limits apply separately: 300 requests per minute for text generation, or 20 for paid-only models on standard billing.