Keep your Gemini app useful through the next model change

Check one task people already rely on

If an app you use or are building runs on Gemini 2.5 Pro or Gemini 2.5 Flash, ask for a test of its replacement before December 1. Pick a familiar job, such as turning an approved sample request into a useful reply. Keep the expected result beside it.

Google announced the change in its November 2 Gemini Developer API release notes. An API is the connection an app uses to ask another service for help. This notice concerns apps using that connection, rather than a change to a personal Gemini subscription.

Read the announcement: Google: November 2 Gemini Developer API retirement notice.

What changes now and on December 1

Gemini 2.5 Pro and Flash are now marked for retirement and are no longer available to select in Google AI Studio. Google says requests sent to those model names will switch automatically on December 1, 2026.

Requests for Gemini 2.5 Pro will go to Gemini 3.8 Flash. Requests for Gemini 2.5 Flash will go to Gemini 3.5 Flash-Lite. Google asks builders to update the model name in their app before that date and follow the matching migration guide.

An automatic switch can change the answers people receive. Treat it as an app update worth checking, even if your existing model name still connects successfully.

Give the connection a proper check

Google lists changes to request settings, conversation formats and the way AI asks an app to carry out an action. Changing a name alone may leave older assumptions in place.

If your app sets how much effort Gemini spends working out an answer, its builder needs to replace the old setting with the newer one. The migration guides also describe older answer settings to remove and conversation formats that can produce an error. Apps that let AI take actions need their own checks against the updated rules.

If you are learning with an AI coding tool, ask it to explain which parts of the official guide apply to your app before it makes changes. Read its explanation, then test a separate copy with sample information. If someone builds for you, ask them to demonstrate the same familiar task before and after the update.

Compare the cost of a completed job

The redirected requests will be billed at the replacement model rates. On the standard paid tier, Gemini 3.8 Flash currently lists $0.75 for a million input tokens and $3.75 for a million output tokens. Tokens are the small pieces of information a model processes. Google says those rates become $1.50 and $7.50 on January 1, 2027.

Gemini 3.5 Flash-Lite lists $0.30 for a million input tokens and $2.50 for a million output tokens on the same tier. Extra tools can add charges. The final cost also depends on how much information your task uses and how much the model produces.

Record the cost of finishing your sample job, including any retries. A lower price for each piece of information does not promise a lower bill for the whole task.

Make this a useful learning exercise

Keep a short record that you can repeat when another service changes.

1. Confirm the model name and the service your app actually calls.
2. Write down what a successful result should contain.
3. Run the replacement with approved sample information and check any actions it takes.
4. Compare the answer, waiting time and full task cost, then agree when the update will be released.

Keep ownership of what good looks like

Our perspective is that knowing how to check a useful result is a lasting skill as AI services change. You can practise it without learning every setting. Start with the task you understand, keep your examples and ask for a clear explanation of any difference that matters to the people using your app.