JA EN
shiichan

AI Search welcomes four new Workers AI models!

Cloudflare Changelog ★★ View Markdown
#cloudflare#ai#serverless

Hey there, it's me, Shiichan! Today I found a fun little update: AI Search just got more models to choose from. Let me walk you through it.

Cloudflare Changelog developers.cloudflare.com

What was announced?

According to the Cloudflare Changelog, AI Search added four more Workers AI models: two for text generation and two for embedding.

The story so far

AI Search ingests your data so you can use it for search and RAG. Inside it, you pick a model that writes text and an embedding model that turns text into vectors. More options mean you can match the length of your documents and your speed preferences more easily.

What changes

All four new models run on Workers AI, so you can use them without bringing any extra provider keys. Just select them when you create or update an AI Search instance in the dashboard or through the API.

Dive Deep

First, the two text generation models.

  • @cf/zai-org/glm-4.7-flash: a lightweight model from Zhipu AI with a 131,072 token context window, a good fit for long-document summarization and retrieval.
  • @cf/qwen/qwen3-30b-a3b-fp8: a mixture-of-experts model from Alibaba with a 32,000 token context. It activates only 3 billion parameters per forward pass, so it stays fast while keeping quality high.

Next, the two embedding models. Both use cosine similarity.

  • @cf/qwen/qwen3-embedding-0.6b: 1,024-dimension vectors and up to 4,096 input tokens, great for indexing longer chunks.
  • @cf/google/embeddinggemma-300m: Google's embedding model producing 768-dimension vectors, with a 512 token input limit and a focus on low-latency workloads.

For the full list, check out Supported models.

Wrap-up

  • AI Search added four Workers AI models: two for text generation, two for embedding.
  • Text generation gets long-context GLM-4.7-Flash and the fast MoE Qwen3-30B-A3B.
  • Embedding gets long-chunk Qwen3-Embedding-0.6B and low-latency EmbeddingGemma-300M.
  • No extra keys needed, just pick them in the dashboard or API. A small but handy update for anyone building RAG or search on Cloudflare!

Related posts