Strands Decider 2B v1 — A Hyperscaler Open-Sources a Calibrated Decision Model (Weights + Data + Recipe, Apache-2.0)
Decider is a 2B model (rank-16 LoRA on top of Qwen3.5-2B-Base; the LM head replaced with a ~1-million-parameter pointer head) that does NOT generate text: given a "state" plus a typed question (`choice` pick 1 of N; `noul` yes/no; `score` ordinal scale), it returns calibrated probabilities per option…
Decider is a 2B model (rank-16 LoRA on top of Qwen3.5-2B-Base; the LM head replaced with a ~1-million-parameter pointer head) that does NOT generate text: given a "state" plus a typed question (`choice` pick 1 of N; `noul` yes/no; `score` ordinal scale), it returns calibrated probabilities per option in a single forward pass — median ~115 ms on an RTX 3090 (vendor claim).
The v1 checkpoint improves on v21: 246,678 rows of data (~2×), adding code tasks and "apply written policy to a case" (executing written rules on a case — exactly the policy-check pattern), and removing 20 thousand rows from licensed dat…