Strands Decider 2B
Amazon Web Services released Strands Decider 2B, an open source 2B decision model that picks from fixed options and scores its confidence, and runs locally.
It is built on the Qwen3.5-2B language model with text generation removed, so it returns a calibrated choice from a closed set of answers instead of text. It follows TypeSafe's Jev design and is released through Strands Labs; the project was started by AWS distinguished engineer Marc Brooker. AWS says it is meant for workflow steps in agents that do not need a full LLM, with lower latency and cost.
- Date
- Thursday 1 October 2026
- Lab
- Amazon
- Kind
- open-weights
- Access
- open weights
Figures
| Measure | Value | Measured by |
|---|---|---|
| Median decision latency on one Nvidia RTX 3090 | about 115 ms | SQ Magazine report |
| JevBench rank in size class (accuracy and calibration combined) | 3rd of 33 models | SQ Magazine report |
Weights, training data and scripts are reported to be on GitHub and Hugging Face; the license was not stated in the pages read.
Sources
- techcrunch.com/2026/10/01/amazon-releases-its-own-jev-clone-as-decision-models-flood-the-w
- sqmagazine.co.uk/amazon-strands-decider-2b-open-source-decision-model/
- pluang.com/en/news-feed/amazon-rilis-model-keputusan-terbuka-jev-klon
This record was checked and corrected against its sources on 8 October 2026. How we check