Cloudflare makes AI Search generally available with native image embeddings and OCR

Cloudflare has made AI Search generally available, combining Workers AI, Vectorize, R2 and Browser Run into a managed index and retrieval pipeline. The service has been in use for over a year, powering search on Cloudflare’s own blog and developer documentation.

The general availability release adds native image embeddings through the Qwen3-VL-Embedding model, optical character recognition for scanned PDFs, and support for larger files. Text files and PDFs up to 10 MiB are now accepted, up from 4 MiB. Image queries are embedded directly by multimodal models, while text-only models fall back to captioning the query image first.

Billing starts on 1 November 2026, with a free monthly tier on all Workers plans that includes 1,000 semantic queries, 1,000 full-text queries and a pool of 5 million ingestion tokens. Cloudflare said it is working on video and audio ingestion, a refactored keyword search engine and simpler ways to create indexes for sites already on its network.