SunsetDiff

Gemini API model shutdowns on November 17, 2026

The dataset has 5 Gemini API ids with shutdown date 2026-11-17: gemini-2.5-flash-native-audio-preview-12-2025, gemini-2.5-flash-preview-tts, gemini-2.5-pro-preview-tts, gemini-3.1-flash-live-preview, gemini-3.1-flash-tts-preview. The deprecations page lists that day as the shutdown date and says the table date is the earliest possible retirement. The two Live audio ids store gemini-3.8-live. The three TTS ids store gemini-3.8-flash-tts and gemini-3.8-flash-lite-tts. The page writes those two TTS ids with or, and does not prefer one. Once a model is shutdown, the endpoint is no longer available.

Last reviewed . Publisher: SunsetDiff.

Steps

  1. Read the five ids. The table is every Gemini API id in the dataset whose shutdown date is 2026-11-17. The replacement cell is the recommended replacement stored from the deprecations page. Open the model page before you change a call.
  2. Split Live audio from text-to-speech. gemini-3.1-flash-live-preview and gemini-2.5-flash-native-audio-preview-12-2025 are Live audio ids. The other three ids are text-to-speech. The speech page says TTS and the Live API are different. A Live replacement does not take a TTS request.
  3. Change the model id and the TTS request. For a Live id, the deprecations page names gemini-3.8-live. For a TTS id, it names gemini-3.8-flash-tts or gemini-3.8-flash-lite-tts. The speech page says Gemini 3.8 TTS treats input text as a verbatim transcript and returns WAV on a unary request.
  4. Test before the date. Call the replacement with one real prompt before November 17, 2026. The deprecations page says it will communicate the exact shutdown date with advance notice. The date in the table is the earliest day it prints.

Models shutting down

These rows are every Google Gemini API model in the dataset with shutdown date 2026-11-17. The model id links to its SunsetDiff model page. Deprecation date, shutdown date, and replacement are the fields stored on that row. Not published means the field is empty. These rows do not store a deprecation_date. The deprecations page lists each of them under Preview models.

The deprecations page says: "Once a model is shutdown, it is completely turned off, and the endpoint is no longer available." It also says: "The shutdown dates listed in the table indicate the earliest possible dates on which a model might be retired. We will communicate the exact shutdown date to users with advance notice to ensure a smooth transition to a replacement model." SunsetDiff stores the printed day, 2026-11-17, as the shutdown date. The page does not print a later day.

The models page says: "A preview name can be used in production and is deprecated with at least two weeks of notice." That sentence is the notice length for a preview name. It does not replace the day in the table.

The dataset has no Vertex row with this shutdown date.

5 Gemini API model ids with shutdown date 2026-11-17.
Model idDeprecation dateShutdown dateRecommended replacement
gemini-2.5-flash-native-audio-preview-12-2025Not published2026-11-17gemini-3.8-live
gemini-2.5-flash-preview-ttsNot published2026-11-17gemini-3.8-flash-tts, gemini-3.8-flash-lite-tts
gemini-2.5-pro-preview-ttsNot published2026-11-17gemini-3.8-flash-tts, gemini-3.8-flash-lite-tts
gemini-3.1-flash-live-previewNot published2026-11-17gemini-3.8-live
gemini-3.1-flash-tts-previewNot published2026-11-17gemini-3.8-flash-tts, gemini-3.8-flash-lite-tts

Sources: Gemini API deprecations , Gemini API models , Vertex model versions , checked November 1, 2026.

Text-to-speech models

The TTS ids in this shutdown are gemini-2.5-flash-preview-tts, gemini-2.5-pro-preview-tts, gemini-3.1-flash-tts-preview. The deprecations page lists gemini-3.8-flash-tts or gemini-3.8-flash-lite-tts for each of them. It does not prefer one.

The speech generation page says: "Use Gemini 3.8 Flash-Lite TTS (gemini-3.8-flash-lite-tts) as your fast, cost-efficient workhorse replacement for gemini-3.1-flash-tts-preview." That sentence names gemini-3.1-flash-tts-preview. It does not print the same sentence for the two Gemini 2.5 preview TTS ids. The migration heading on that page says it covers gemini-3.1-flash-tts-preview or earlier Gemini TTS models.

The same page says: "Both Gemini 3.8 TTS models share the exact same API schema and prompting format, allowing you to switch between them with a single parameter change." And: "TTS models accept text-only inputs and produce audio-only outputs." And: "Gemini 3.8 TTS treats input text strictly as a verbatim transcript." And: "Unlike gemini-3.1-flash-tts-preview and earlier TTS models (which returned headerless raw PCM audio/l16 by default), Gemini 3.8 TTS returns WAV audio (audio/wav) with a standard RIFF header by default for unary requests." If you need headerless raw PCM, the page says to set response_format.mime_type to audio/l16, audio/mulaw, or audio/alaw.

The speech page's samples call client.interactions.create and POST https://generativelanguage.googleapis.com/v1beta/interactions. The model field in those samples is gemini-3.8-flash-tts. The page also documents gemini-3.8-flash-lite-tts.

Sources: Gemini API deprecations , Gemini speech generation , checked November 1, 2026.

Live audio models

The Live audio ids in this shutdown are gemini-2.5-flash-native-audio-preview-12-2025, gemini-3.1-flash-live-preview. The deprecations page lists gemini-3.8-live for each of them.

The speech page says: "The TTS capability differs from speech generation provided through the Live API, which is designed for interactive, unstructured audio, and multimodal inputs and outputs." A TTS model id is not the replacement for these Live rows.

The gemini-3.8-live model page says: "Gemini 3.8 Live is the default option for most low-latency voice agent experiences and real-time dialogue without reasoning-induced delays." It says the model code is gemini-3.8-live, and it points to the Live API guide for features. This guide does not copy a Live request. The Live API guide is the page for that shape.

Sources: Gemini API deprecations , Gemini speech generation , gemini-3.8-live model page , Gemini Live API , checked November 1, 2026.

What breaks

On November 17, 2026 these ids reach the shutdown date stored in the dataset. The deprecations page says: "Once a model is shutdown, it is completely turned off, and the endpoint is no longer available." It does not print one HTTP status for that moment.

The same page says: "The shutdown dates listed in the table indicate the earliest possible dates on which a model might be retired. We will communicate the exact shutdown date to users with advance notice to ensure a smooth transition to a replacement model." The day in the table can move if Google sends a later exact date. Until that notice, the stored shutdown date stays 2026-11-17.

A 400 from a field the replacement rejects is a different failure. The speech page's WAV and speech_metadata changes are request-shape changes on Gemini 3.8 TTS, not the shutdown itself.

Sources: Gemini API deprecations , Gemini speech generation , checked November 1, 2026.

Find the model ids in your code

Search the repository for the five ids in the table. A preview suffix is part of the id. gemini-2.5-flash is a different id from gemini-2.5-flash-preview-tts. How to find every model ID in your codebase covers SDK argument names. The scan page builds a search from dataset ids that have a shutdown date.

Also search configs that pick a voice model by a short label. The deprecations page uses the full id.

ripgrep

rg -n -w -e 'gemini-2.5-flash-native-audio-preview-12-2025' -e 'gemini-2.5-flash-preview-tts' -e 'gemini-2.5-pro-preview-tts' -e 'gemini-3.1-flash-live-preview' -e 'gemini-3.1-flash-tts-preview'

Source: Gemini API deprecations , checked November 1, 2026.

Test the replacement before the cutoff

Call the replacement id from the same row with the prompts you send now. Do that before November 17, 2026. For TTS, the speech page says the two Gemini 3.8 TTS models share one API schema, so you can switch between gemini-3.8-flash-tts and gemini-3.8-flash-lite-tts by changing that one field. Run the one you will keep, and read the audio bytes. A unary response is WAV unless you set another mime type.

For the Live ids, call gemini-3.8-live from the Live API path you use now. The model page says Live API is supported on that id. This guide does not name a generateContent body for it.

The pre-migration test checklist is the longer list: evals, rejected fields, token limits, tools, price, and latency.

Sources: Gemini API deprecations , Gemini speech generation , gemini-3.8-live model page , checked November 1, 2026.

Related SunsetDiff pages

All guides

Questions

Which Gemini API ids shut down on November 17, 2026?

The dataset lists gemini-2.5-flash-native-audio-preview-12-2025, gemini-2.5-flash-preview-tts, gemini-2.5-pro-preview-tts, gemini-3.1-flash-live-preview, gemini-3.1-flash-tts-preview with shutdown date 2026-11-17. The deprecations page says: "The shutdown dates listed in the table indicate the earliest possible dates on which a model might be retired. We will communicate the exact shutdown date to users with advance notice to ensure a smooth transition to a replacement model."

What is the recommended replacement?

The recommended replacement is the official replacement stored on each row. gemini-2.5-flash-native-audio-preview-12-2025: gemini-3.8-live. gemini-2.5-flash-preview-tts: gemini-3.8-flash-tts, gemini-3.8-flash-lite-tts. gemini-2.5-pro-preview-tts: gemini-3.8-flash-tts, gemini-3.8-flash-lite-tts. gemini-3.1-flash-live-preview: gemini-3.8-live. gemini-3.1-flash-tts-preview: gemini-3.8-flash-tts, gemini-3.8-flash-lite-tts. The deprecations page writes or between the two TTS ids and does not prefer one. The speech page says: "Use Gemini 3.8 Flash-Lite TTS (gemini-3.8-flash-lite-tts) as your fast, cost-efficient workhorse replacement for gemini-3.1-flash-tts-preview."

What does Google say happens at shutdown?

The deprecations page says: "Once a model is shutdown, it is completely turned off, and the endpoint is no longer available." It does not print one HTTP status for that moment.

What changes in a TTS request besides the model id?

The speech page says: "Gemini 3.8 TTS treats input text strictly as a verbatim transcript." And: "Unlike gemini-3.1-flash-tts-preview and earlier TTS models (which returned headerless raw PCM audio/l16 by default), Gemini 3.8 TTS returns WAV audio (audio/wav) with a standard RIFF header by default for unary requests." It also says: "TTS models accept text-only inputs and produce audio-only outputs." Samples on that page set model to gemini-3.8-flash-tts and POST https://generativelanguage.googleapis.com/v1beta/interactions.

Are the Live ids the same migration as the TTS ids?

No. gemini-2.5-flash-native-audio-preview-12-2025, gemini-3.1-flash-live-preview store gemini-3.8-live. The speech page says: "The TTS capability differs from speech generation provided through the Live API, which is designed for interactive, unstructured audio, and multimodal inputs and outputs." The gemini-3.8-live model page says: "Gemini 3.8 Live is the default option for most low-latency voice agent experiences and real-time dialogue without reasoning-induced delays."

Does this page include Vertex models?

No Vertex row shares this shutdown date.

More guides

Deprecation alerts

We email you before models shut down. Unsubscribe anytime.

Deprecation alert signup
Providers (optional)