●DEPRECATION — The gemini-omni-flash-preview endpoint retires on September 30. Its replacement, gemini-omni-1.1-flash, reached general availability on August 27, so this one needs attention this month●MIGRATION — The swap is often a single model string, but the default resolution is now 720p. Carry over your preview-era assumptions unchanged and both your output and your bill will shift●OMNI — Version 1.1 closed four gaps: scene extension out to 40 seconds, first-and-last-frame control, a cheap 360p draft mode, and 4K upscaling●COPILOT — Gemini 3.8 Flash became selectable in GitHub Copilot on September 3, widening its reach a day after arriving in the Gemini API, AI Studio, and Antigravity●PRICE — The $0.75 input and $3.75 output per MTok on 3.8 Flash is introductory. It expires December 31, and from January 1, 2027 the rate becomes $1.50 and $7.50 per MTok●SPEECH — Gemini 3.5 Transcribe is a pair of dedicated speech-to-text models, with utterance-level language detection across 85+ languages, speaker diarization, word-level timestamps, and custom vocabulary biasing●DEPRECATION — The gemini-omni-flash-preview endpoint retires on September 30. Its replacement, gemini-omni-1.1-flash, reached general availability on August 27, so this one needs attention this month●MIGRATION — The swap is often a single model string, but the default resolution is now 720p. Carry over your preview-era assumptions unchanged and both your output and your bill will shift●OMNI — Version 1.1 closed four gaps: scene extension out to 40 seconds, first-and-last-frame control, a cheap 360p draft mode, and 4K upscaling●COPILOT — Gemini 3.8 Flash became selectable in GitHub Copilot on September 3, widening its reach a day after arriving in the Gemini API, AI Studio, and Antigravity●PRICE — The $0.75 input and $3.75 output per MTok on 3.8 Flash is introductory. It expires December 31, and from January 1, 2027 the rate becomes $1.50 and $7.50 per MTok●SPEECH — Gemini 3.5 Transcribe is a pair of dedicated speech-to-text models, with utterance-level language detection across 85+ languages, speaker diarization, word-level timestamps, and custom vocabulary biasing