serve_decide.py: System 1 only by default for every model (thinking is opt-in)

serve_decide.py: one file for JEV-27B and GEV; optional adaptive thinking (thinking=auto|on, reasoning_effort, think_budget; off by default for JEV-27B), answer read-out that also works with speculative decoding

Model card: "New (1 October 2026)" summary at the top

System 1 over HTTP (POST /v1/decide, serve_decide.py), choice up to 256 options, prompt guide, 256K context test

Update README.md

Publish JEV-27B model and evaluation results