The Workers AI team at Cloudflare has released two decision models named clef designed for high accuracy and low latency. These internally trained models are available through Cloudflare's Workers AI hosting platform and as open source weights on Hugging Face, which allows developers to run the tools outside of the provider's environment.

This launch marks the first time the network security company has trained and published its own AI models rather than deploying third party architectures. According to the release, clef currently tops performance benchmarks for quality and speed, signaling a shift toward in-house development of specialized AI tools.