Amazon's AI Model: Fast, Cheap Decisions

Amazon Web Services recently unveiled Strands Decider 2B, an open-source AI model inspired by TypeSafe's Jev. This new offering is designed to provide AI developers with intelligence optimized for computer automation, rather than the extensive capabilities of frontier LLMs. The Strands Decider 2B is engineered for high-speed, low-cost decision-making among pre-defined options. A key feature is its ability to deliver confidence scores for its choices, thereby increasing the reliability of automated workflows.

Developed by Amazon distinguished engineer Marc Brooker, the project originated from observing customer needs for more streamlined AI agentic workflows. The model, based on the Qen3.5-2B LLM's core, functions by outputting calibrated choices instead of generating text. Its small size allows it to run locally, offering reduced latency and potentially lower costs for specific agent tasks. The open-source release aims to democratize access to efficient decision-making tools for AI development.