Workers AI is a serverless inference service from Cloudflare, Inc. Inference means running a trained AI model on new input to get an output; serverless means that the customer does not rent, manage or scale the GPUs (graphics processors) that run the models. The models run on GPUs in Cloudflare's network, in more than 200 cities.
The catalog has more than 50 open-source models for tasks such as text generation, image generation, embeddings, image classification and speech. Code calls a model by name in one call: from a Cloudflare Worker (a small program that runs on Cloudflare's servers), through the REST API from any language, or through the OpenAI SDK, because the API is OpenAI-compatible. For speech, the catalog includes OpenAI's Whisper, a speech recognition model that turns audio into text, translates speech and identifies the language.
Workers AI is a hosted, pay-per-use service. Use is measured in Neurons, a unit for the GPU compute that a request needs, and a free allocation of Neurons each day is included.