[{"data":1,"prerenderedAt":7},["ShallowReactive",2],{"docs-changelog\u002F2026-11-07":3,"docs-tabs-changelog\u002F2026-11-07":6},{"content":4,"lastmod":5},"# API changes: November 7, 2026\n\n[← Changelog](\u002Fdocs\u002Fchangelog) · [November 2026](\u002Fdocs\u002Fchangelog\u002F2026-11)\n\n### FIX-1107-1: Lifecycle headers for transcription models\n\n**Before**\n\n[POST \u002Fv1\u002Faudio\u002Ftranscriptions](\u002Fdocs\u002Fai\u002Faudio\u002Ftranscriptions) ignored the model status and its replacement, although the [model lifecycle rules](\u002Fdocs\u002Fai\u002Fmodels\u002Flifecycle) also apply to speech recognition.\n\n**After**\n\nThe endpoint follows the documented rules: a deprecated model returns `Deprecation`, `Sunset` (when a shutdown time is set), and `Link` \u002F `X-Model-Replacement` (when a replacement is available). An available replacement serves requests to a disabled model with `X-Model-Fallback`, `X-Model-Replacement`, and `X-Model-Fallback-Reason: model_disabled` headers. Only a disabled model without an available replacement returns `503 model_unavailable`.\n\n**Impact on integrators**\n\nNo integration changes are required. The headers let integrators track model retirement and identify the model serving a request under the shared lifecycle rules.\n",null,{},1794739486759]