Google shuts down Gemini 2.5 Pro and Flash in its API on 1 December. What happens to your automations?

Is Google shutting down Gemini 2.5 Pro and Gemini 2.5 Flash?

Yes. Google's Gemini API deprecations page, read on 6 November 2026, gives a shutdown date of 1 December 2026 for both gemini-2.5-pro and gemini-2.5-flash. That is a change of course. In a changelog entry dated 18 September 2026, Google limited the 2.5 models to people who had already used them, but said they were not deprecated and would be served until further notice.

Google also says gemini-2.5-pro can no longer be selected in Google AI Studio, and the 2.5 text-to-speech and Live previews shut down sooner, on 17 November 2026. These dates are for the Gemini API. If you reach Gemini through Google Cloud's enterprise platform, check that platform's own model pages, because the Gemini API pages do not give dates for it.

Will my automations stop working on 1 December?

Mostly no, and that is the risk. Google says calls to the old model names will be routed to the new models automatically. A Zapier, Make or n8n step, a website chatbot or a script set to gemini-2.5-flash will keep returning answers, but from a different model, with different defaults and a different bill. Nobody gets an error message telling them to look.

Google's migration guides list the changes that matter:

• Temperature stops working. Google says the API ignores temperature, top_p and top_k on these models, and that future models will return an error if you send them. If you set a low temperature to keep answers consistent, that setting no longer does anything.
• Thinking settings change. The guides replace thinking_budget with thinking_level. Gemini 3.5 Flash-Lite defaults to "minimal" thinking. Gemini 3.8 Flash defaults to "medium" and returns an error if you ask for "minimal".
• Prefilled replies fail. If the last turn in a request is a model turn, a trick some prompts use to start the answer, the API returns a 400 error.

What does the switch cost?

Here are Google's paid prices per 1 million tokens, read on 6 November 2026:

Two things stand out. Gemini 2.5 Pro work gets cheaper in December, but 3.8 Flash's input price doubles on 1 January, to more than 2.5 Pro charged. Jobs that read long documents and write short answers may cost more from January. Gemini 2.5 Flash work keeps the same list price but moves to a Lite model. Google's own guide says minimal thinking suits extraction, routing and classification, and that steps with tool calls or multi-step reasoning should use "medium" or "high" to stop the model ending tool use too early.

How do I find every Gemini 2.5 step in my business?

Work through this checklist this week:

1. 1Check your Gemini API usage and billing for calls to gemini-2.5-pro and gemini-2.5-flash.
2. 2Open every automation with a Google Gemini step in Zapier, Make and n8n, and write down the model each one uses.
3. 3Search your code and settings for "gemini-2.5-pro" and "gemini-2.5-flash". A search for "gemini-2.5" also finds Flash-Lite, which is not affected.
4. 4List third-party tools that hold your Gemini API key, such as chat widgets and plugins, and check their model setting.
5. 5Note the settings each step sends: temperature, thinking budget or a prefilled answer. These are what Google's guides change.
6. 6Mark each step as simple or judgement work. Sorting and tagging is simple. Tool calls, agents and anything written in your brand voice need closer testing.

If the list runs past a handful of steps, our AI workflows service starts with exactly this kind of workflow map and can take the switch off your plate.

Should you let Google reroute you or switch yourself?

Our advice is to switch yourself before 1 December, so you choose the model and its settings on purpose. Simple 2.5 Flash steps can move to gemini-3.5-flash-lite at minimal thinking, which is the job Google describes it for. Flash steps that call tools should be tested on 3.5 Flash-Lite at medium thinking and compared with gemini-3.8-flash. Gemini 2.5 Flash-Lite is still available, but our advice is not to build anything new on the 2.5 family.

Worked example: a ten-person online shop with three Gemini steps

This is a hypothetical scenario, not a client.

1. 1Day 1: inventory. The operations lead finds an n8n workflow that sorts customer emails with gemini-2.5-flash, a Make scenario that writes product descriptions with gemini-2.5-pro at temperature 0.4, and a website chat widget on gemini-2.5-flash that looks up orders with a tool call.
2. 2Day 2: copy and set. She duplicates each one. The email sorter moves to gemini-3.5-flash-lite at minimal thinking. The description writer moves to gemini-3.8-flash with the temperature removed. The chat widget gets two copies: 3.5 Flash-Lite at medium thinking and 3.8 Flash.
3. 3Week 1: test. She runs 20 real past inputs through old and new versions. If the chat widget answers before it looks up the order, she raises its thinking level, as Google's guide suggests.
4. 4Week 2: brand check. Temperature no longer keeps the descriptions consistent, so she adds two approved descriptions to the prompt as examples and reads ten new ones side by side with the old.
5. 5Before 24 November: switch. Live steps move to the models she chose. The old versions stay paused for a week.
6. 6December and January: watch the bill, especially after 1 January, when 3.8 Flash input prices double.

What other AI model deadlines are coming?

Gemini is not the only one. OpenAI shuts down GPT-4o and GPT-4.1 in its API on 14 January 2027 (see our guide to the GPT-4o shutdown), and OpenAI Agent Builder closes on 30 November (alternatives here). Our advice: keep one list of every AI model your business calls, with its shutdown date, and check it monthly.

Can Theepa help?

Yes. Our AI workflows and automation service maps the workflows you have, moves them to the right models and leaves a setup your team can run.

Sources

• Gemini API deprecations (Google AI Studio docs, accessed 6 November 2026)
• Gemini API changelog, entry of 18 September 2026 (Google AI Studio docs, accessed 6 November 2026)
• Gemini models (Google AI Studio docs, accessed 6 November 2026)
• Gemini Developer API pricing (Google AI for Developers, accessed 6 November 2026)
• What's new in Gemini 3.8, migration guide (Google AI Studio docs, accessed 6 November 2026)
• What's new in Gemini 3.6, Flash-Lite migration and API changes (Google AI Studio docs, accessed 6 November 2026)