Cloudflare Workers AI

Cloudflare Workers AI lets developers run machine-learning models on Cloudflare's global network without managing GPUs. You call models from Workers, Pages, or any app through a REST API. The catalog holds more than 50 open models for text, images, audio, and embeddings. Billing uses Neurons, a GPU-compute unit, with 10,000 free Neurons per day. It also offers OpenAI-compatible endpoints, so existing OpenAI code works with small changes. Fine-tuning with LoRA adapters and tools like AI Gateway round out the platform.

Key Features

1. Serverless GPU inference

Models run on Cloudflare's network with no servers to manage. Scaling and hardware are handled for you.

2. 50+ model catalog

Pick from open models for text generation, classification, detection, and embeddings. New models are added over time.

3. OpenAI-compatible API

Endpoints match the OpenAI format, so you can switch providers by changing a URL. Existing client libraries keep working.

4. Daily free Neurons

Every account gets 10,000 Neurons per day at no charge. Paid usage costs $0.011 per 1,000 Neurons after that.

5. LoRA fine-tuning

Adapt models with lightweight LoRA adapters, including public ones. Custom behavior needs far less data than full training.

6. AI Gateway and Vectorize

AI Gateway adds caching, rate limits, and fallbacks to model calls. Vectorize provides a vector database for search and RAG.

Frequently Asked Questions (FAQ)

How does Neuron billing work?

Neurons measure GPU compute per request. You get 10,000 free per day, resetting at midnight UTC. Extra usage needs the Workers Paid plan.

Can I use my OpenAI code with it?

Yes. The OpenAI-compatible endpoints accept the same request format. Point your client at the new base URL and keep your code.

Which models need a paid plan?

Some frontier models, like certain Kimi, GLM, and DeepSeek versions, require a paid billing method. The catalog marks which models those are.

Is Workers AI still in beta?

No. It is generally available. Function calling and the batch API remain in beta while they mature.

Conclusion

Workers AI is a practical way to add AI to apps already on Cloudflare. The daily free Neurons cover real prototyping, and serverless scaling removes ops work. The OpenAI-compatible API makes trying it nearly free in engineering time. If your stack lives on Cloudflare, it is the obvious first place to run models. Start with the docs quickstart.