JA EN
shiichan

AI Search Welcomes a Super-Long-Context Model: GLM-5.3 Flash!

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

Hi, I'm Shii-chan! Cloudflare's AI Search just welcomed another new model, and this one has a seriously impressive spec: a context window of 1,048,576 tokens. Let's take a look!

Cloudflare Changelog developers.cloudflare.com

What was announced?

According to Cloudflare's Changelog, AI Search now supports a new text generation model, @cf/zai-org/glm-5.3-flash (GLM-5.3 Flash). It's built by Zhipu AI (zai-org) and runs on Workers AI. The context window is 1,048,576 tokens, and you can find out how to configure it for an AI Search instance in Cloudflare's "Supported models" documentation.

The story so far

AI Search already offered a GLM model from zai-org called GLM-4.7 Flash, which had a 131,072-token context window. GLM-5.3 Flash multiplies that by 8x, letting you feed in up to 1,048,576 tokens at once. That should cut down a lot on the chunking work needed when you want to search or summarize long documents.

What changes

If you're building RAG (retrieval-augmented generation) with AI Search, you can now just pick GLM-5.3 Flash from the model settings to get a model that handles a much longer context. That's especially useful if you want to hand over a big pile of internal docs or logs at once, or have the model answer questions across multiple files.

Dive Deep

I checked out the GLM-5.3 Flash model page too, and found a few more interesting details.

  • It's the first natively multimodal model in the GLM-5 series, supporting function calling, reasoning, and vision
  • It has 320B total parameters, with only 18B active parameters thanks to a mixture-of-experts (MoE) design
  • It reportedly outperforms the previous GLM-5.2 on benchmarks and real-world workloads, at one-tenth the price
  • On coding and agentic benchmarks, it's said to approach Claude Opus 4.8
  • Pricing is $0.15 per million input tokens, $0.50 per million output tokens, and $0.03 per million cached input tokens
  • It's not available on the Workers Free plan; you'll need the Workers Paid plan or prepaid AI Gateway credits

Wrap-up

  • AI Search now supports a new text generation model, GLM-5.3 Flash (@cf/zai-org/glm-5.3-flash)
  • Its context window is 1,048,576 tokens, 8x the previous GLM-4.7 Flash (131,072 tokens)
  • It's a multimodal model with function calling, reasoning, and vision, and is said to approach Claude Opus 4.8 on coding and agentic benchmarks
  • You'll need the Workers Paid plan or prepaid AI Gateway credits to use it
  • If you're building RAG and want to feed in whole long documents, this is a nice new option to have!

Related posts