AWS open-sourced a 2B model that only makes decisions
AWS’s Strands Labs released Strands Decider 2B on Wednesday, an open-source decision model built by taking the torso of Qwen3.5-2B, removing the language-model head that lets it generate text, and replacing it with a small pointer head that scores the answers offered to it. The head adds just over a million parameters and the torso is fine-tuned with a rank-16 LoRA. It is the nineteenth iteration of the design, and everything — weights, training data and scripts — is on Hugging Face and GitHub.
Decision models, or “system one” models, answer only by choosing among options and scoring how sure they are, which makes them fast and cheap where an LLM call would be wasteful. Strands says the model ranks third of 33 in the two-billion-parameter class — first of 30 once the just-over-2B entries are excluded — on JevBench’s accuracy and calibration, and decides in a median of about 115 milliseconds on an RTX 3090 and 153 on an M3 MacBook. Its worked example is an agent that asks whether a tool call is grounded in what the user actually said before making it.
WHY IT MATTERS
AWS is betting that a whole class of agent decisions — routing, tool selection, guardrails, policy checks — does not need a frontier model and should not pay for one, and the class it is chasing was opened by TypeSafe AI’s Jev earlier this month. For a builder the appeal is structural rather than benchmark-driven: a decision that costs milliseconds and runs on a laptop is one you can afford to put in front of every tool call, which is exactly where the failure modes live. The more valuable half is the blueprint in the repo — training a decider on your own task is a weekend, not a research program. The caveat deserves one line: the accuracy is measured on JevBench’s public set, and the model cannot generate text at all.