Overview
Workers AI is Cloudflare's serverless inference service. It runs a catalogue of open-weight models, including Llama, Mistral, DeepSeek and IBM Granite for text, plus image and other task types (50+ open-source models), on GPUs in Cloudflare's network. You call it from a Worker through a binding or from anywhere through a REST API.
Billing uses Neurons, a unit of GPU compute, with published per-model equivalents per million tokens: Llama 3.2 1B costs $0.027 in and $0.201 out, Llama 3.1 70B $0.293 / $2.253, and DeepSeek V4 Pro $1.32 / $3.96. Every account gets 10,000 Neurons a day free, on both the Free and Paid plans.
On the Free plan, requests beyond the daily allocation fail with an error rather than being billed. On Workers Paid, usage above 10,000 Neurons a day costs $0.011 per 1,000 Neurons. This makes it a low-risk way to add small-model features to a Workers app.
Pricing and plans
Free tier · no card
10,000 Neurons per day at no charge on both the Workers Free and Paid plans.
At the limit Stops at the limit On the Free plan, operations beyond 10,000 Neurons a day fail with an error until the daily allocation resets.
| Plan | Price | What you get |
|---|---|---|
| Workers Free | $0 | 10,000 Neurons/day; requests fail beyond that |
| Workers Paid | $0.011 per 1,000 Neurons | After 10,000 free Neurons/day |
| Llama 3.2 1B | $0.027 in / $0.201 out per 1M | |
| Mistral 7B | $0.110 in / $0.190 out per 1M | |
| Llama 3.1 70B | $0.293 in / $2.253 out per 1M | |
| DeepSeek V4 Pro | $1.320 in / $3.960 out per 1M |
What it does
- Daily free allocation
- 10,000 Neurons a day on both Free and Paid plans.
- Neuron billing
- $0.011 per 1,000 Neurons beyond the free allocation on Workers Paid, with per-model token equivalents.
- Workers binding
- Call models directly from Cloudflare Workers code, or through a REST API.
- Open-weight catalogue
- 50+ open-source models for text generation, image classification, object detection and more.
Strengths and limitations
Strengths
- Free daily allocation that resets and never bills on the Free plan.
- Runs next to Workers code on the same platform and bill.
- Very cheap small models, e.g. Llama 3.2 1B at $0.027 per 1M input tokens.
Limitations
- Only open-weight models from Cloudflare's catalogue; no proprietary frontier models.
- Neuron-based billing is less intuitive than plain per-token pricing.
- Beyond 10,000 Neurons a day you need the Workers Paid plan.
- Large-model prices are not lower than specialist providers (DeepSeek V4 Pro at $1.32 / $3.96).
Who it suits
Good for
- Apps already built on Cloudflare Workers
- Small AI features such as summaries or classification
- Hobby projects that must never incur a bill
Look elsewhere if
- Apps that need frontier proprietary models
- Heavy LLM workloads where a dedicated inference provider is cheaper
Alternatives to Cloudflare Workers AI
| Tool | From | Free option | Stages |
|---|---|---|---|
| SambaNova Cloud (SambaCloud) Inference provider (open-weight models) | Usage-based | Free tier | |
| Novita AI Inference provider (open-weight models) | Usage-based | ||
| OpenRouter Model router (unified API) | — | Free tier | |
| DeepInfra Inference provider (open-weight models) | Usage-based | ||
| Gemini API AI model API (model maker) | Usage-based | Free tier | |
| GroqCloud Inference provider (open-weight models) | Usage-based | Free tier |
Questions
Is Cloudflare Workers AI free?
Every account gets 10,000 Neurons a day free on both Free and Paid plans. On the Free plan, operations beyond that fail with an error.
What is a Neuron in Workers AI?
Cloudflare's unit for the GPU compute a model uses; beyond the free allocation it costs $0.011 per 1,000 Neurons on Workers Paid.
How much does Llama 3.1 70B cost on Workers AI?
$0.293 per 1M input tokens and $2.253 per 1M output tokens on 2026-09-29.
Sources read for this page