Amazon releases open-source decision model inspired by Jev

Amazon Web Services has launched Strands Decider 2B, an open-source decision model that aims to provide efficient and cost-effective automated decision-making. This release comes in the same week that OpenAI unveiled a similar offering, highlighting a growing trend in AI development towards specialized models suited for automation rather than general-purpose large language models (LLMs).

Strands Decider 2B is built upon the Qen3.5-2B LLM but is designed to output calibrated choices with confidence scores, rather than generating text. This approach allows for lower latency and potentially reduced costs for specific workflow steps. The model is small enough to be run locally and is fully open-sourced.

The project was initiated by Amazon distinguished engineer Marc Brooker, who was inspired by TypeSafe's Jev model. Brooker observed a need among AWS customers for AI tools that could handle specific decision-making tasks within agentic workflows without the full capabilities and associated costs of frontier LLMs. He noted that the confidence scores and closed domain of answers contribute to more reliable workflows.

TypeSafe, the company behind the Jev model, views the proliferation of similar models as a sign of wide interest but questions their true value. CEO Diogo Almeida suggested that many current implementations focus on architecture rather than deep intelligence, and that making AI truly smart is a significant challenge. He indicated that TypeSafe is focused on improving its future models and does not currently see direct competition emerging.