AWS Open Sources Strands Decider 2B as Jev Rivals Multiply
AWS has open sourced Strands Decider 2B, a local model built on Qwen3.5-2B that promises faster, cheaper, confidence scored choices for AI agent workflows.
Summary
AWS unveiled Strands Decider 2B on October 1, 2026, an open source decision model inspired by TypeSafe’s Jev and released the same week as a similar OpenAI offering. Available now and small enough to run locally, it rapidly selects among predefined options and provides calibrated confidence scores, giving AI agent workflows a potentially cheaper, faster and more reliable alternative to using frontier LLMs for every step. It uses the LLM torso of Qwen3.5-2B but returns choices instead of generated text.
Amazon distinguished engineer Marc Brooker began the project after seeing Jev; his prototype briefly led the Jevbench rankings for models of its size before AWS engineers prepared it for Strands Labs, which develops AI agent deployment tools and protocols. AWS customers had sought simpler workflow decision tools without full LLM costs or capabilities. Dozens of Jev-like models now exist, with some costing hundreds or thousands of dollars to build, leading Brooker to doubt frontier labs will dominate smaller markets. The challenge is increasing speed, accuracy and calibration without weakening language understanding or knowledge. TypeSafe, which named Jev after economist William Stanley Jevons and his theory that lower costs can increase demand, is improving future models. CEO and founder Diogo Almeida says current rivals have not yet matched the difficulty of making intelligence genuinely useful.
Positives
- Strands Decider 2B is fully open source, available now and compact enough to run locally.
- Calibrated confidence scores and predefined answers could make agent workflow steps more reliable.
- The prototype briefly ranked first on Jevbench among models of its size.
- Hundreds or thousands of dollars can fund development, opening the field beyond frontier AI laboratories.
- Qwen3.5-2B provides general language understanding while the model returns fast, structured choices.
Risks & concerns
- Dozens of similar models raise doubts about differentiation and lasting commercial value.
- Higher decision speed could undermine accuracy, calibration, multilingual understanding or retained knowledge.
- OpenAI launched a similar offering in the same week, adding immediate competition.
- TypeSafe CEO Diogo Almeida says current rivals have not demonstrated genuinely useful intelligence.
- Smaller target markets may constrain the opportunity even if development costs remain low.