Cloudflare Unifies Workers AI and AI Gateway into Single Control Plane
Cloudflare is merging its Workers AI and AI Gateway products into a unified control plane, simplifying access to AI models, enhancing observability, and enabling unified billing.
Cloudflare has announced a significant consolidation of its AI offerings, merging the Workers AI and AI Gateway products into a single, unified control plane. This strategic move aims to streamline the developer experience by providing a consistent interface for accessing various AI models, regardless of their underlying provider.
Previously, Workers AI offered managed model hosting on Cloudflare's GPU infrastructure, while AI Gateway acted as a proxy for external model providers, offering features like observability, logging, and security. The convergence means developers can now interact with any model, including those hosted on Workers AI, through a single REST API or a unified Workers binding. This eliminates the need to choose between distinct products, presenting a more cohesive entry point for AI integrations.
A key benefit of this unification is the automatic provision of observability and control for all Workers AI users. Previously, users had to explicitly set up an AI Gateway to gain visibility into their inference traffic. Now, by simply specifying 'default' as the gateway ID in their API calls or bindings, users automatically inherit AI Gateway's features. This includes comprehensive logging of requests and responses, token usage tracking per model, and cost attribution, all without requiring manual dashboard configuration.
For users who require more advanced customization, the option to create named gateways remains. This allows for features like custom caching rules or traffic splitting across different applications. The transition to a named gateway is designed to be seamless, requiring only a single parameter change in the API call or binding configuration.
Furthermore, Cloudflare is introducing unified billing for Workers AI usage. Developers can now use AI Gateway credits to pay for inference on Workers AI models, alongside external providers like OpenAI and Anthropic. This consolidated billing approach simplifies financial management and is complemented by elevated rate limits for Workers AI models when accessed through the unified AI Gateway billing system, encouraging adoption of the new path.
Looking ahead, Cloudflare plans to introduce "model-first routing." This future capability will shift the focus from selecting a specific model provider to specifying the desired model's capabilities. The AI Gateway will then intelligently handle provider selection, load balancing, and failover, ensuring resiliency and optimal performance. For instance, a developer could request a specific model like "Kimi K2.7 Code" without needing to know which provider hosts it, with the gateway transparently routing the request to the most suitable available option.
This evolution towards a unified control plane signifies Cloudflare's commitment to simplifying the complexities of AI deployment. By abstracting away infrastructure management and providing integrated tools for observability, security, and billing, the company aims to empower developers to build and scale AI-powered applications more efficiently and reliably.