Gemini API Drops 3.7 Flash for 3.8 Flash: What Pakistani Developers Must Change

Google deprecated two Gemini API models on 8 October 2026. Calls to gemini-3.7-flash now go to gemini-3.8-flash, and calls to gemini-3.5-flash now go to gemini-3.6-flash. The older Deep Research agent, deep-research-pro-preview-12-2025, shuts down on 23 October 2026.

If your Pakistani startup, agency or freelance project calls the Gemini API with either old model name, your app has not broken. Google now routes those requests to the newer model on its own. But the answers your app gets back now come from a different model, and that can change output length, format and cost.

We read Google's Gemini API release notes, its deprecations table and its pricing page on 10 October 2026. Everything below comes from those three pages. We did not run a migration test ourselves.

Google retired Gemini 3.7 Flash and 3.5 Flash on 8 October

On 8 October 2026 Google marked gemini-3.7-flash and gemini-3.5-flash as deprecated in the Gemini API. Both still answer, because Google forwards every request to a replacement model. Google asks developers to update the model string in their code.

A deprecation, in Google's own words, is the notice that a model is no longer supported. A shutdown comes later, when the endpoint is switched off completely. Neither old Flash model has a shutdown date yet. The routing is what matters today.

Gemini 3.7 Flash had only been out since 13 August 2026. It lasted under two months as the main Flash model. That pace is a fair warning for anyone who hard-codes a model name and forgets about it.

The Deep Research preview agent stops on 23 October

The Deep Research agent deep-research-pro-preview-12-2025 shuts down on 23 October 2026, and Google lists no automatic routing for it. After that date, requests using that agent name will fail. You have to change the agent parameter yourself.

Google offers two newer agents. The release note describes them like this:

• deep-research-preview-04-2026: built for speed, and suited to streaming results back to a user screen.
• deep-research-max-preview-04-2026: built for the most complete research and summary work.

The change is one line. In your interactions.create request, swap the old agent name for one of the two new ones. Test a few real research prompts before 23 October, because a different agent may return longer reports or take longer to finish.

Gemini 3.8 Flash pricing doubles on 1 January 2027

Google lists Gemini 3.8 Flash at US$0.75 per million input tokens and US$3.75 per million output tokens on the paid tier through 31 December 2026. From 1 January 2027 the listed prices become US$1.50 and US$7.50. Gemini 3.6 Flash shows the same two price steps.

A token is a small piece of text, roughly part of a word. Output tokens include the model's thinking tokens, so a model that reasons longer costs more on the same prompt.

Here is a simple example. Say a Lahore support chatbot uses 10 million input tokens and 2 million output tokens a month on Gemini 3.8 Flash. At today's price that is US$7.50 plus US$7.50, so US$15. From January the same traffic costs US$30. Your card is charged in dollars, so the rupee bill also moves with the exchange rate.

Cheaper options exist. Google lists Gemini 3.5 Flash-Lite at US$0.30 input and US$2.50 output per million tokens, with no price change date shown. Google's Batch API also cuts paid prices by 50% for jobs that can wait. The free tier of the API still lists 3.8 Flash as free of charge, but Google says free-tier content is used to improve its products, and paid-tier content is not.

How to move your app to the new models

Update the model string, test with real prompts, then watch your bill for a week. Routing keeps old code running, but it hides the switch. Making the change yourself means you see any difference in output before your users do.

1. Search your code and settings for gemini-3.7-flash and gemini-3.5-flash.
2. Replace them with gemini-3.8-flash, or gemini-3.6-flash if you want the lower-level option.
3. Search for deep-research-pro-preview-12-2025 and swap in a 04-2026 agent.
4. Run 20 to 30 of your real prompts and compare the answers with old logs.
5. Check output length and format, especially JSON your app parses.
6. Watch token use in Google AI Studio for a week after the change.

Freelancers who built a Gemini tool for a client should tell the client now. A model change is easy to miss, and a client who sees different answers will ask you first. A short note with the new model name and the January price saves that call.

The Gemini app is a separate product. The free Gemini app change, where free users in Pakistan now get Flash-Lite only, is about the chat app, not the API. If you are choosing between chat plans rather than coding, free ChatGPT and Gemini plans in Pakistan is the better comparison.

More Gemini API shutdown dates to note

Several older preview models switch off on 17 November 2026, and these will fail rather than reroute. Most are voice and text-to-speech previews. Gemini 3.1 Flash-Lite has a later shutdown date of 7 May 2027.

If you run an Urdu voice bot or a call centre tool on a Live preview model, 17 November is your real deadline.

Common questions

Will my app stop working because of the 8 October change?

No, not for the two Flash models. Google routes gemini-3.7-flash to gemini-3.8-flash and gemini-3.5-flash to gemini-3.6-flash. The Deep Research preview agent is different and stops on 23 October 2026.

Do I need to change my API key?

No. The change is only to the model name or agent name in your request. Your key and billing stay the same.

Is Gemini 3.8 Flash more expensive than 3.7 Flash?

Google's pricing page no longer lists 3.7 Flash, so we cannot compare directly. It lists 3.8 Flash at US$0.75 input and US$3.75 output per million tokens until 31 December 2026, doubling from 1 January 2027.

Can I still use Gemini 3.8 Flash free from Pakistan?

Google's pricing page shows a free tier for 3.8 Flash with lower rate limits. Free-tier content can be used to improve Google's products, so keep client or customer data on the paid tier, in line with PKCERT's advice on using AI chatbots safely at work.

Which Gemini model costs the least for simple text work?

Gemini 3.1 Flash-Lite is listed at US$0.25 input and US$1.50 output per million text tokens, but it shuts down on 7 May 2027. Its listed successor, Gemini 3.5 Flash-Lite, costs US$0.30 input and US$2.50 output. The Batch API halves paid prices for work that can wait.

How we verified this

What we checked, where we read it, and what we could not confirm.

We last checked this on 10 October 2026. We read Google's Gemini API release notes (8 October 2026 entry), the Gemini API deprecations table and the Gemini API pricing page. Prices are in US dollars per million tokens as Google lists them, before any bank or card charges in Pakistan.