Amazon launches open-source AI model designed to make fast decisions

Amazon Web Services has released Strands Decider 2B, an open-source artificial intelligence model designed to make fast, low-cost decisions within automated AI workflows.

Unlike conventional large language models that generate open-ended text, Strands Decider is designed to select between predefined options, rate outcomes and provide a confidence score for its decisions.

The model was inspired by Jev, a decision model developed by TypeSafe, as developers increasingly explore smaller systems that can handle routine AI-agent decisions without relying on larger, more expensive frontier models.

What does Strands Decider 2B do?

The model is aimed at tasks that sit between traditional classification systems and full large language models.

Potential uses include:

• Choosing which AI model should handle a task
• Selecting the next tool an AI agent should use
• Routing customer requests to the correct team
• Checking whether AI-generated actions meet certain requirements
• Scoring outputs
• Applying safety or policy classifications

Rather than writing a full response, the model evaluates available choices and returns a structured decision.

Small enough to run locally

Strands Decider 2B has around 1.9 billion parameters, making it significantly smaller than many frontier AI systems.

AWS says the model can run on devices including Apple silicon Macs and CPUs, while benchmark testing recorded median response times of around 115 milliseconds on an RTX 3090.

This makes it suitable for workflows where speed and cost matter more than generating complex text.

Built from Qwen3.5-2B

The model uses Qwen3.5-2B-Base as its underlying foundation.

However, instead of retaining the language-generation component found in a normal LLM, the system replaces it with a specialised decision-making layer designed to score available choices.

This allows it to make calibrated decisions without going through a full text-generation process.

Project started inside Amazon

Amazon distinguished engineer Marc Brooker developed the initial version after experimenting with the concept behind TypeSafe’s Jev.

The project briefly reached the top of the JevBench rankings for models of its size, leading Amazon engineers to refine it and release it through Strands Labs, AWS’s experimental open-source initiative for agentic AI development.

AWS customers wanted cheaper AI decisions

Brooker said the idea was influenced by discussions with AWS customers whose AI-agent workflows did not always require the capabilities — or cost — of a full large language model.

For many automated processes, an AI system may simply need to decide what action to take next, rather than generate a detailed response.

That creates an opportunity for smaller decision models to handle routine steps while larger models are reserved for more complex tasks.

OpenAI is exploring a similar approach

AWS is not alone in moving towards specialised decision models.

OpenAI recently introduced its Decisions API, which similarly allows developers to provide a predefined set of options for an AI system to evaluate and choose between.

The product was announced in limited preview and is designed to make certain automated AI decisions faster and more efficient.

The growing interest in these systems reflects a broader shift in AI development: instead of using increasingly powerful large models for every task, developers are exploring combinations of large reasoning models and smaller specialised models to reduce latency and operating costs.

About Qatar Living:

Since 2005, Qatar Living has been the trusted destination for everything Qatar. As the country's largest online community and marketplace, Qatar Living connects people with opportunities through jobs, property, vehicles, services, classifieds, events, local insights, and breaking news. Trusted by residents, newcomers, visitors, and businesses alike, Qatar Living brings Qatar together in one place.

Instagram - @qatarliving

X - @qatarliving

Facebook - Qatar Living

YouTube - qatarlivingofficial