JA EN
shiichan

A New Model Just Landed on Workers AI: Meet GLM-4.7-Flash

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

Hey everyone, it's me! Cloudflare dropped three AI updates at once today, so let me walk you through all of them.

Cloudflare Changelog developers.cloudflare.com

What was announced?

According to the Cloudflare Changelog, this release bundles three things together.

  • A new Workers AI model, @cf/zai-org/glm-4.7-flash
  • A new package for TanStack AI, @cloudflare/tanstack-ai
  • An update to the Vercel AI SDK provider, workers-ai-provider, now at v3.1.1

They all share one theme: making it easier to build agents (AI that calls tools across multiple turns) entirely on Cloudflare's edge.

Why it matters

Building an agent usually means juggling several pieces at once: picking a model, wiring up a framework, and making sure the SDK you use plays nicely with everything. This time, the model, the framework integration, and the SDK all got updated together, so Workers AI now covers most of the agent-building stack on its own. That's worth knowing about.

What changes

GLM-4.7-Flash is described as a fast, efficient text generation model optimized for multilingual dialogue and instruction-following. The changelog highlights these strengths:

  • Multi-turn tool calling, built for agents
  • Multilingual support
  • A large context window
  • Fast inference
  • Strong instruction-following

You can call it the same way you'd call any other Workers AI model, through:

  • The Workers AI binding (env.AI.run())
  • The REST API (/run or /v1/chat/completions)
  • AI Gateway
  • The Vercel AI SDK, via workers-ai-provider

@cloudflare/tanstack-ai is a new framework-agnostic package that brings Workers AI and AI Gateway support to TanStack AI. It covers chat with tool calling, image generation, transcription, text-to-speech, and summarization, and it supports both a plain binding setup and an AI Gateway binding (env.AI.gateway(id)). If you're already using TanStack, connecting to Workers AI just got a lot simpler.

Finally, workers-ai-provider v3.1.1 improves the experience for anyone using the Vercel AI SDK.

Dive Deep

The v3.1.1 update to workers-ai-provider covers two areas.

New capabilities

  • Speech-to-text transcription
  • Text-to-speech generation
  • Document reranking for RAG pipelines

Reliability improvements (building on v3.0.5)

  • Token-by-token streaming with proper backpressure handling
  • Fixes for tool-calling issues
  • Better stream-termination detection

If you've ever had a stream cut off mid-response or a tool call get missed, these fixes should make a real difference, even if they're not flashy.

Installing everything follows the usual npm pattern:

npm install @cloudflare/tanstack-ai @tanstack/ai
npm install workers-ai-provider@latest ai

One thing to note: the changelog doesn't list specific pricing for GLM-4.7-Flash. You'll need to check the model page or Cloudflare's pricing page for exact numbers.

Wrap-up

  • Workers AI added a new multilingual model, @cf/zai-org/glm-4.7-flash, with a 131,072-token context window
  • @cloudflare/tanstack-ai is a new official package covering chat, image generation, transcription, text-to-speech, and summarization for TanStack AI
  • workers-ai-provider v3.1.1 adds transcription, text-to-speech, and document reranking, plus more reliable streaming
  • You can reach the model through env.AI.run(), the REST API (/run or /v1/chat/completions), AI Gateway, or the Vercel AI SDK

If you're building agents on Cloudflare's edge, or you're already using TanStack or the Vercel AI SDK, this update is worth checking out right away.

Related posts