--- title: AWS releases Strands Decider 2B, an open decision model... description: AWS's Strands Labs released Strands Decider 2B, a small open-source decision model built on Qwen3.5-2B and inspired by TypeSafe's Jev. Rather than generating... canonical: https://daily.dev/posts/aws-releases-strands-decider-2b-an-open-decision-model-modeled-on-typesafe-s-jev-w1ctygq0t twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: AWS releases Strands Decider 2B, an open decision model modeled on TypeSafe's Jev | daily.dev og:description: AWS's Strands Labs released Strands Decider 2B, a small open-source decision model built on Qwen3.5-2B and inspired by TypeSafe's Jev. Rather than generating... og:url: https://daily.dev/posts/aws-releases-strands-decider-2b-an-open-decision-model-modeled-on-typesafe-s-jev-w1ctygq0t og:image: https://api.daily.dev/og/posts/w1CTYgq0t.png og:image:alt: AWS releases Strands Decider 2B, an open decision model modeled on TypeSafe's Jev og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # AWS releases Strands Decider 2B, an open decision model modeled on TypeSafe's Jev **[Collections](https://daily.dev/sources/collections)** · 2 min read · 2 upvotes · 0 comments ## Summary AWS's Strands Labs released Strands Decider 2B, a small open-source decision model built on Qwen3.5-2B and inspired by TypeSafe's Jev. Rather than generating text, it scores developer-supplied answer options to give AI agents a fast grounding check before acting, with latency under 100ms on an RTX 3090 and about 150ms on an M3 MacBook. It integrates with AWS's Strands agent framework via an intervention system that can let an action proceed, deny it, ask for confirmation, or return feedback, while generative calls still route through Amazon Bedrock. AWS published training data and scripts and claims the top spot on JevBench among models with a full public training recipe. The release lands amid a crowded week of competing decision models, including OpenAI's hosted Decisions API, Kev, imajev, and Laya, with TypeSafe's CEO downplaying the competitive threat. ## Content AWS's Strands Labs has released Strands Decider 2B, an open-source "decision model" inspired by TypeSafe's Jev. It is built on Qwen3.5-2B and doesn't generate free text. It scores answer options the developer supplies and returns a confidence value. The idea is to give an AI agent a fast, cheap grounding check before it acts. ## How it works The model is a small download that runs locally. AWS reports sub-100ms latency on an RTX 3090 and around 150ms on an M3 MacBook. It says the model ranks first on JevBench among public models that come with a full training recipe, and it has published the training data and scripts. It plugs into AWS's Strands agent framework through an intervention system. A decision can let the agent proceed, deny the action, ask for confirmation, or return feedback. Generative calls still go through Amazon Bedrock. ## A crowded week The release landed the same week OpenAI announced its hosted Decisions API. Decider also competes with Kev, imajev, and Laya. @thdxr reposted @ryanvogel, who put OpenAI's API up against Jev and Clef and called it "FASTTT." AWS engineer Marc Brooker built the original homebrew version, which briefly topped the JevBench leaderboard for its size class. He says the hard part from here is balancing speed against keeping general intelligence. TypeSafe's CEO Diogo Almeida downplays the competition. In his view, most new entrants copy the architecture without the depth needed to make it useful. ## Why not an encoder? Brooker also tested whether an encoder-only, bidirectional design like BERT would beat the modified decoder-only LLM behind Decider. He compared five designs, including T5Gemma variants, ModernBERT, and ettin-encoder-1b, on accuracy, calibration, and latency, using JevBench and JF100. The results were mixed: - Encoders won on latency with shorter prompts but scaled worse because of quadratic attention costs. - The decoder-based v19 model, which uses Gated DeltaNet linear-cost layers, held up on accuracy and generalized better than an unmasked encoder variant. - Brooker acknowledges the experiment is confounded across several variables. At this model size, no architecture comes out a clear winner. ## Questions this post answers ### What is AWS's Strands Decider 2B and how does it work with AI agents? Strands Decider 2B is a small open-source decision model from AWS's Strands Labs, built on Qwen3.5-2B and inspired by TypeSafe's Jev. Instead of generating free text, it scores developer-supplied answer options and returns a confidence score, giving agents a fast grounding check before acting. It plugs into AWS's Strands agent framework through an intervention system that can let an action proceed, deny it, ask for confirmation, or request feedback. _Teams wiring guardrails into agent pipelines can follow decision-model releases like this on daily.dev._ ### How fast is Strands Decider 2B compared to generative LLM calls for agent decisions? Strands Decider 2B runs locally with latency under 100ms on an RTX 3090 and around 150ms on an M3 MacBook, since it only scores pre-decided options rather than generating text. Generative calls in the same pipeline still route through Amazon Bedrock, so the decision model acts as a cheaper, faster gate before a full generative call is made. _Developers benchmarking agent latency can track model comparisons like this on daily.dev._ ### What is JevBench and why does Strands Decider 2B rank first on it? JevBench is a benchmark for decision models, the category pioneered by TypeSafe's Jev. AWS says Strands Decider 2B ranks first on JevBench among public models that ship with a full training recipe, meaning both the model and its training data and scripts are published, unlike some competitors that only release weights. _Those evaluating decision models for agent guardrails can follow benchmark results on daily.dev._ ## Similar posts on daily.dev - [Small Decisions: Engineering a Leading Model](https://daily.dev/posts/small-decisions-engineering-a-leading-model-9wqnf1abu) · Marc Brooker · 2 upvotes · 1 comments --- Tags: [#open-source](https://daily.dev/tags/open-source), [#aws](https://daily.dev/tags/aws), [#ai-agents](https://daily.dev/tags/ai-agents) [View this post on daily.dev](https://daily.dev/posts/aws-releases-strands-decider-2b-an-open-decision-model-modeled-on-typesafe-s-jev-w1ctygq0t) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"AWS releases Strands Decider 2B, an open decision model modeled on TypeSafe's Jev","url":"https://daily.dev/posts/aws-releases-strands-decider-2b-an-open-decision-model-modeled-on-typesafe-s-jev-w1ctygq0t","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/aws-releases-strands-decider-2b-an-open-decision-model-modeled-on-typesafe-s-jev-w1ctygq0t"},"datePublished":"2026-10-01T16:51:28.197Z","dateModified":"2026-10-06T21:45:00.815Z","description":"AWS's Strands Labs released Strands Decider 2B, a small open-source decision model built on Qwen3.5-2B and inspired by TypeSafe's Jev. Rather than generating...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/86f611913a454d95bbec47cc1f7d8816?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/86f611913a454d95bbec47cc1f7d8816?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Collections","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Collections","logo":"https://media.daily.dev/image/upload/s--fk_6ycEi--/f_auto,q_auto/v1780996001/logos/collections?_a=BAMAMiWQ0","url":"https://daily.dev/sources/collections"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/aws-releases-strands-decider-2b-an-open-decision-model-modeled-on-typesafe-s-jev-w1ctygq0t","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":2},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"open-source,aws,ai-agents","timeRequired":"PT2M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Collections","item":"https://daily.dev/sources/collections"},{"@type":"ListItem","position":3,"name":"AWS releases Strands Decider 2B, an open decision model modeled on TypeSafe's Jev"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/aws-releases-strands-decider-2b-an-open-decision-model-modeled-on-typesafe-s-jev-w1ctygq0t#faq","mainEntity":[{"@type":"Question","name":"What is AWS's Strands Decider 2B and how does it work with AI agents?","acceptedAnswer":{"@type":"Answer","text":"Strands Decider 2B is a small open-source decision model from AWS's Strands Labs, built on Qwen3.5-2B and inspired by TypeSafe's Jev. Instead of generating free text, it scores developer-supplied answer options and returns a confidence score, giving agents a fast grounding check before acting. It plugs into AWS's Strands agent framework through an intervention system that can let an action proceed, deny it, ask for confirmation, or request feedback. Teams wiring guardrails into agent pipelines can follow decision-model releases like this on daily.dev."}},{"@type":"Question","name":"How fast is Strands Decider 2B compared to generative LLM calls for agent decisions?","acceptedAnswer":{"@type":"Answer","text":"Strands Decider 2B runs locally with latency under 100ms on an RTX 3090 and around 150ms on an M3 MacBook, since it only scores pre-decided options rather than generating text. Generative calls in the same pipeline still route through Amazon Bedrock, so the decision model acts as a cheaper, faster gate before a full generative call is made. Developers benchmarking agent latency can track model comparisons like this on daily.dev."}},{"@type":"Question","name":"What is JevBench and why does Strands Decider 2B rank first on it?","acceptedAnswer":{"@type":"Answer","text":"JevBench is a benchmark for decision models, the category pioneered by TypeSafe's Jev. AWS says Strands Decider 2B ranks first on JevBench among public models that ship with a full training recipe, meaning both the model and its training data and scripts are published, unlike some competitors that only release weights. Those evaluating decision models for agent guardrails can follow benchmark results on daily.dev."}}]} ```