Cloudflare says its new Clef model means humans no longer need to be in the loop for AI agents
decoder ·
Cloudflare has launched Clef and Clef-flash, two new decision models designed to enable AI agents to operate without human intervention. Unlike traditional large language models that generate text, these models assign probabilities to predefined answer options, allowing downstream systems to act automatically on the results. Cloudflare positions these tools as direct competitors to TypeSafe AI's Jev model, emphasizing significantly faster response times of 39 milliseconds for Clef-flash compared to over 500 milliseconds for its rival. Built on Qwen architectures, the models support both text and image inputs and feature a 64,000-token context window. The company reports that Clef achieves higher decision quality while maintaining speed advantages across various benchmarks. Additionally, Cloudflare is introducing a reinforcement learning service that allows customers to fine-tune the models for specific tasks using their own data. These models are available on Hugging Face under an Apache-2.0 license and run on Cloudflare's Workers AI platform.
decoder