Strands Decider 2B
AWS Strands Labs' tiny decision model — 1.9B params, fully open, runs on a consumer GPU.
About
Strands Decider 2B, released October 1, 2026 by AWS Strands Labs, is the small end of the decision-model wave: roughly 1.9 billion parameters built on Qwen3.5-2B, with the language-model head removed and replaced by a scoring head of around a million parameters. It is released openly — weights, training data, and training scripts — under Apache 2.0.
The pitch is self-hosting: it is small enough to run on a consumer GPU, with a reported median latency around 115 ms on an RTX 3090. Its own launch post concedes it is significantly worse than reasoning models on complex problems — an honest and useful statement of what this category is for: fast, cheap, structured judgments, not deliberation.
It launched the same day as Cloudflare's Clef pair and Perplexity's pplx-decider, aimed at the same classification-and-routing use cases TypeSafe's Jev opened in September.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-reported: median ~115 ms on RTX 3090; concedes worse than reasoning models on complex tasks. Independent: none.
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Release facts (October 1, 2026; ~1.9B on Qwen3.5-2B; scoring-head architecture; Apache 2.0 with weights, data, and scripts; ~115 ms on RTX 3090) reported consistently across independent coverage of the launch wave. Docked for: (1) no first-party model card or benchmark table located — details come via secondary coverage; (2) no independent evaluation; (3) the vendor itself positions it well below reasoning models on hard tasks.