AI that plans, uses tools and works through problems step by step.
An AI agent does not answer and stop. It plans, calls tools, reads the results and decides what to do next, repeating until it judges the task finished. That loop is what makes coding assistants, research tools and computer-using systems possible, and it is also where these systems fail in new ways: looping, misusing a tool, or stopping confidently with the wrong answer. This page follows agents and the reasoning methods they depend on — the releases, the benchmarks that try to measure them, and the honest accounts of what still does not work.
Most AI models reply in sentences. Cloudflare's new Clef models pick an answer from a fixed list and say how sure they are, which suits jobs like sending a support ticket to the right team. There are two versions: Clef, built on a 27-billion-parameter Qwen model, and the smaller Clef-flash, built on a 9-billion one. Both are free to download under the Apache 2.0 licence. Cloudflare says a typical answer takes 209.3 ms from Clef and 38.8 ms from Clef-flash.