Strands Decider 2B
Strands Agents · Released 1 Oct 2026
A small, open-source 2-billion-parameter decision model for fast experimentation and local development. Built for picking among predefined options and assigning simple numerical scores, with a reliability score for each decision. Intended for agent workflows such as model routing, tool selection, evaluations, and policy classification.
- Maker
- Strands Agents
- Release date
- 1 Oct 2026
- First seen
- 1 Oct 2026
- Access
- Open source on GitHub, with weights on Hugging Face, including training data and scripts. A strands-decider CLI is provided. Can run on a local CPU or GPU.
- Pricing (as publicly stated)
- Open source; self-hosted. No hosted list price was stated in the announcement.
- Modalities
- text
- Question types
- choice, noul, score
- Latency (as stated)
- The team reported a median of around 115 ms on a local NVIDIA RTX 3090, and around 153 ms for small tasks on an M3 MacBook.
- Official announcement
- https://strandsagents.com/blog/introducing-strands-decider/
Benchmark claims
These are claims as published by the maker or in the linked official post. We have not independently rerun every benchmark.
The authors said accuracy and calibration on JevBench's public set placed the model 3rd of 33 in the 2B class, and 1st of 30 excluding just-over-2B models, and that it performed 100% of the easy tasks on JevBench correctly.
Names and marks belong to their owners. This page is an independent summary, not a product page from Strands Agents.