AWS announced Strands Decider 2B on October 1, 2026: an open-source model that scores a set of possible answers for a specific decision, such as which tool an AI agent should use. Developers can inspect and modify its weights, code, training data and scripts, and run it locally. That makes openness and self-hosting its clearest distinction from TypeSafe AI’s hosted Jev API; the available comparisons do not show that Strands Decider is better overall.
What Strands Decider does
Rather than write an unrestricted response, a decision model evaluates choices supplied by an application and returns a selection or scores. AWS describes uses such as classifying text, choosing a category, routing a request, selecting a tool, evaluating an output or classifying a policy. It can also answer multiple questions about the same prompt.
For example, AWS’s launch post poses the bounded question, “Is the string ‘turn on the lights’ about the coffee machine? Yes or no.” Another asks which language the phrase “sihamba ngokushesha” is in, offering English, Zulu or Dutch as choices. The application defines the choices; the model judges among them.
AWS presents this as a component for a workflow, not a replacement for a general-purpose generative model. A generative model can handle complex reasoning or produce language, while a decider handles a defined choice point. AWS says decision models are significantly worse at complex problems than reasoning models and are unsuitable for coding, chatbots and document summarization.
Free tools Windows power users keep installed
One-click scans. No signup required.
What AWS released and how it works
The Strands Agents team says Strands Decider 2B starts from Qwen3.5-2B. AWS removed the base model’s language-model head and replaced it with a pointer head that scores answer options against a representation of the question. The team fine-tuned the model’s torso using a rank-16 LoRA adapter; AWS describes the pointer head as just over one million parameters. The release includes model weights, code, training data and scripts. AWS’s Strands Decider documentation describes the model and its use.
The launch post calls the released checkpoint version 19, following earlier design iterations. AWS says it can run on a local CPU or GPU. In one example, it is used with Strands’ agent-intervention mechanism before a tool call: the decider assesses whether proposed tool arguments are grounded in what the user said and whether clarification should come first. The questions and thresholds in that example were chosen manually. The model supplies a judgment; the developer still sets the policy and decides what the application does with the result.
How it compares with Jev
Strands Decider resembles Jev in its focus on making bounded decisions, but the evidence supports a narrow comparison rather than a winner. Strands is released for inspection, adaptation and self-hosting; Jev is accessed as a hosted API. Those differences matter to developers weighing local control and privacy against the work of operating a model themselves.
| Comparison point | Strands Decider 2B | Jev |
|---|---|---|
| Access described in launch coverage | Open-source release with weights, code, training data and scripts, according to AWS. Source | Hosted API, according to VentureBeat’s comparison. Source |
| Local operation | AWS says it can run on a local CPU or GPU. Source | Not stated in the cited comparison. Source |
| Evidence that one is more accurate overall | Not established: the chart discussed by VentureBeat does not include Jev. Source | Not established by the cited comparison. Source |
| Comparable operating cost or latency | Not established: AWS did not provide a general self-hosting cost estimate, and hosted Jev and local Strands latency were measured under different conditions. Source | Not established on comparable conditions in the cited comparison. Source |
A useful comparison would need matched workloads and hardware, as well as accuracy, calibration, latency and total cost—including compute and maintenance. The cited evidence does not show that Strands is categorically faster, more accurate or cheaper, or that it is a “Jev killer.”
What the launch benchmarks show—and what they do not
AWS reports that Strands Decider ranked third among 33 models in the 2B class on JevBench’s public set, and first among 30 after models slightly above 2B parameters were excluded. These are AWS-reported results for that benchmark set, not a general measure of performance across applications.
AWS also reports median latency of around 115 milliseconds on its cited local hardware and around 153 milliseconds for small tasks on an M3 MacBook. The latency graph uses an RTX 3090 to measure an earlier checkpoint, v18; AWS says latency grows approximately linearly with task size. These figures describe the stated tests, not a guarantee for other hardware, input lengths or workloads.
VentureBeat reads AWS’s chart as showing roughly 72% accuracy and a 0.35 Brier score for v19, compared with about 76% accuracy and a 0.32 Brier score for Mapika’s similarly sized model. A Brier score measures probability calibration, not accuracy; the two figures describe different qualities. Jev is absent from the chart, so it cannot establish how Strands performs against Jev.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Why decision models are attracting attention
TechCrunch reports that AWS distinguished engineer Marc Brooker began the project after seeing Jev and building his own version; AWS later released the work through Strands Labs. Brooker described the role as a workflow step that decides “what is the next thing for me to do here, based on where I am?” TechCrunch says researchers had produced “dozens” of similar models after TypeSafe introduced Jev, but that is not a sourced census or formal measure of the market. Read TechCrunch’s report.
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteA September 30, 2026 arXiv preprint examines Jev for recommendation reranking across Amazon Reviews domains and candidate-set sizes. Its abstract reports strong recommendation effectiveness relative to the study’s tested baselines and more gradual latency growth than pointwise Qwen rerankers, while finding Jev slower than recommendation-specific models. That is evidence about one recommendation task, not a comparison with Strands or a verdict on decision models generally. Read the preprint abstract.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
