Cloudflare is merging its AI Gateway and Workers AI products into one unified control plane that automatically routes your requests to the best available model provider, eliminating the need to manage multiple AI infrastructure vendors separately. Starting today, developers can request specific models like Kimi K2.7 Code without specifying which provider hosts it, and the system will automatically load balance across Workers AI, Moonshot's own API, or other vetted providers if one becomes unavailable.

Why Should You Care About Model-First Routing?

Traditionally, developers have had to think like infrastructure managers. You pick a provider, OpenAI or Anthropic, and if that provider goes down or hits rate limits, your application breaks. Cloudflare's new approach flips this on its head. Instead of asking "Which provider do I call?", you ask "What kind of task am I solving?", and the control plane handles the rest.

This matters because it solves a real operational headache. If you're building an application that needs a capable reasoning model, you can now request Kimi K2.7 Code and get it from whichever provider has capacity at that moment. If Workers AI is at capacity, the gateway transparently routes you to another provider hosting the same model weights. The quality remains consistent because Cloudflare works only with vetted providers and respects requirements like Zero Data Retention.

What New Features Come With the Unified Control Plane?

The convergence brings several practical improvements for developers already using either product. Every Workers AI user now gets automatic observability and logging without any setup required. When you make a request, the system logs full request and response payloads, tracks token counts per model, and attributes costs automatically.

Cloudflare has also introduced unified billing across all supported providers. Previously, you could only use AI Gateway credits on external providers like OpenAI and Anthropic. Now, you can load a wallet full of credits and spend them across OpenAI, Anthropic, Workers AI, or any other supported provider in a single billing system.

How to Get Started With the Unified Control Plane

• For Workers AI Users: Route your existing calls through a default gateway by adding a third argument to your binding or REST API call. You'll immediately get request logging, token tracking, and cost attribution without changing anything else about how you call the models.
• For AI Gateway Users: Add Workers AI to your existing setup by calling a Workers AI model directly. Load up your AI Gateway wallet and you'll get unified billing across every provider, plus elevated rate limits on Workers AI models.
• For New Users: Set up your first gateway, browse the Workers AI model catalog, and start building. The system creates a default gateway automatically on your first authenticated request if you've never set one up before.

The technical implementation uses a single AI binding and a unified REST API endpoint called "/ai/" that works for both Workers AI and AI Gateway. This means developers no longer need to choose between products; everything comes with observability and control built in.

What About Intelligent Routing?

Cloudflare is piloting an even more advanced feature that goes beyond simple failover. The company is building intelligent routing that understands what you're asking for and picks the right model automatically. A classifier running on Workers AI reads your prompt and predicts what kind of task it is, whether it's coding, research, summarization, or general question-answering, and how complex it is. A heuristic scorer then maps that to the best model from a curated pool.

This zero-config path means teams get better economics and performance without maintaining their own routing logic. For teams that want control, you can still specify exact models. Cloudflare is currently piloting this internally and will actively test and iterate in the coming weeks before release.

The shift toward model-first routing represents a fundamental change in how developers think about AI infrastructure. Instead of managing provider relationships and handling failover logic in application code, the control plane becomes the intelligent middleman. This approach also improves resiliency by default; if one provider's version of a model has issues, traffic shifts to another without requiring application-level retries or complex fallback logic in your Workers.

For developers already invested in either AI Gateway or Workers AI, the unification is designed to be frictionless. You can start small with a default gateway and scale up to named gateways with custom caching rules and traffic splitting as your needs grow. The unified billing system means you're no longer locked into a single provider's pricing or availability constraints.