Cloudflare AI Search: General Availability With Native Image Search, OCR, and Larger File Support
Cloudflare AI Search exits preview, adding direct image pixel embeddings for visual search, built-in OCR for scanned PDFs, increased file size limits (10 MiB), and updated pricing with a new free tier—all available now, with billing to start November 1, 2026.
What changed?
Cloudflare AI Search is now generally available. Major updates include the ability to embed image pixels directly for visual search (with Qwen3-VL-Embedding), native support for optical character recognition (OCR) on scanned PDFs, and support for ingesting larger files (up to 10 MiB, up from 4 MiB). The API now allows seamless image and text search, supporting both types with native multimodal embeddings. Free usage now provides 1,000 semantic and 1,000 full-text queries per month per user. Pricing is finalized, but billing begins only on November 1, 2026.
Why does it matter to an everyday developer?
This release offers immediate, out-of-the-box multimodal search capabilities on Cloudflare Workers. Developers can now index and search not only documents and web resources, but also images and scanned PDF content—greatly broadening use-cases to product catalogs, document search, media discovery, and knowledge base applications where visual similarity and OCR matter. Larger file limits reduce the need to chunk sources manually. Updated pricing and a generous free tier make it easy to test or run small apps at no cost before billing starts in late 2026.
What can the developer do now?
- Enable AI Search in the Cloudflare dashboard or via API to create search indexes.
- Ingest image files and scanned PDFs (up to 10 MiB each), taking advantage of built-in OCR and image embedding.
- Choose the Qwen3-VL-Embedding model for native multimodal visual search, or use text models for fallback caption-based search.
