apply_openai_reasoning_effort matches muse-spark-1.1/1.2/1.2-contributor only, so 1.3 receives zero reasoning_effort; extend the family match (prefix, not per-release literals) so catalog-discovered 1.3 inherits the dialect. CODEX_PICKER_EFFORTS lacks XHigh/Ultra rows though the wire already maps both to xhigh with an xhigh display label; add the rows so GPT-5.6-class models expose them. Tests: wire values per tier, picker lists per route.
apply_openai_reasoning_effort matches muse-spark-1.1/1.2/1.2-contributor only, so 1.3 receives zero reasoning_effort; extend the family match (prefix, not per-release literals) so catalog-discovered 1.3 inherits the dialect. CODEX_PICKER_EFFORTS lacks XHigh/Ultra rows though the wire already maps both to xhigh with an xhigh display label; add the rows so GPT-5.6-class models expose them. Tests: wire values per tier, picker lists per route.