Amazon releases Strands Decider 2B and Cloudflare launches Clef as small open 'decision models' multiply
The models pick among fixed options and return confidence scores, which their makers pitch as cheap, reliable steps in agent workflows.
Amazon Web Services has released Strands Decider 2B, an open-source decision model small enough to run locally. On the same day, Cloudflare introduced Clef, a set of open-source decision models alongside a new reinforcement-learning fine-tuning platform. TechCrunch says the releases are part of a wave of models inspired by TypeSafe AI's Jev, which launched earlier this month. Decision models differ from general LLMs. Rather than generating arbitrary text, they choose among a set of options, such as whether a request concerns a coffee machine or which language a phrase is in, or assign simple numerical scores. According to the Strands team, this makes them faster and more capable at a given size, guarantees an answer from the allowed options and allows very low latency. The trade-off is that generating all outputs in a single parallel pass makes them much worse than reasoning models at complex problems. Amazon distinguished engineer Marc Brooker started the project as a personal attempt to build his own version of Jev. It briefly topped the Jevbench ranking for models of its size, and AWS's Strands Labs then cleaned it up for release. Brooker told TechCrunch that customers' agent workflows do not always need a full LLM, and that confidence scores and a closed set of answers make a workflow step more reliable. For developers, these models offer a cheaper way to handle routing and classification inside agents.