Gemini 2.5 Flash is Google's Flash-tier model from the 2.5 generation (model code gemini-2.5-flash); as of Nov 2026 Google's developer page says access to the 2.5 models is limited to people who have actively used them in the past, and points new projects to newer models. In the Gemini app itself the choices now are Flash-Lite, Flash, Pro and Argon, and the app picks for you by default.
So there are two answers, depending on where you meet Gemini. If you use the Gemini app, you won't find a 2.5 Flash option at all. If your team or developer calls Google's models through the API, 2.5 Flash is still running, but Google is steering new work elsewhere. Here's what each side needs to know.
What Gemini 2.5 Flash is
Google's developer page describes it as its best model for price against performance, with well-rounded capabilities, and says 2.5 Flash is best for large scale processing, low-latency, high volume tasks that require thinking, and agentic use cases. In plain words: it was built to answer quickly and in bulk, rather than to be the deepest thinker in the range.
The specification on that page, as of Nov 2026:
- Model code: gemini-2.5-flash
- What it reads: text, images, video and audio
- What it writes: text only
- Input token limit: 1,048,576
- Output token limit: 65,536
- Knowledge cutoff: January 2025
- Latest update: June 2025
Supported capabilities include caching, code execution, file search, function calling, grounding with Google Maps, search grounding, structured outputs, thinking and URL context. Audio generation, image generation and the Live API aren't supported. For developers, it can run through the Batch API, Flex inference and Priority inference.
Can you still use it?
This is the part most people searching for it want. Google's page carries a note that says, in short: to keep performance reliable for everyone, Google is limiting access to the 2.5 models to users who have actively used them in the past. The note adds that these models "are not deprecated and will continue to be served until further notice through the API". For any new projects, Google tells you to use its latest models: 3.5 Flash-Lite or 3.8 Flash.
Read that carefully, because it cuts both ways. If something your business already relies on calls gemini-2.5-flash, it isn't being switched off, and there's no shutdown date on the page. But if you're starting something new, Google itself is telling you not to build on it.
One version is gone already. The page lists the stable gemini-2.5-flash as current, and the preview version gemini-2.5-flash-preview-09-2025 as shut down. If an old setup points at that preview name, that's the one to update.
On Google's main models page, Gemini 3.8 Flash is described as "Our most intelligent Flash model", and Gemini 3.6 Flash as the previous-generation Flash model. That's the direction Google is pointing.
Where you meet Gemini, and what you can pick
| Where you use Gemini | What you can pick | What Google's page says |
|---|---|---|
| Gemini app, no Google AI plan | Auto or Flash-Lite | With auto on, most prompts use Flash-Lite and harder ones may use Flash or Pro; with it off, everything uses Flash-Lite |
| Gemini app, Google AI Plus | Flash-Lite, Flash, Pro | Thinking levels Low, Medium and High |
| Gemini app, Google AI Pro or Ultra | Flash-Lite, Flash, Pro, Argon in chats; Flash and Argon in Tasks | Argon is top-tier intelligence with dynamic deep reasoning |
| Gemini with a work or school account | Depends on your Workspace licence | The personal-account details don't apply |
| Gemini API, for developers | gemini-2.5-flash if your project used it before | Not deprecated; new projects should use 3.5 Flash-Lite or 3.8 Flash |
Why the app doesn't mention 2.5
In the Gemini app, the model names don't carry version numbers. Google's help page describes them like this: Flash-Lite is a fast, cost-efficient model for quick answers to everyday questions; Flash is a high-speed model that balances performance and responsiveness; Pro is an advanced reasoning model for complex prompts, sophisticated coding and in-depth problem-solving; and Argon is top-tier intelligence for complex workflows.
So if a tutorial tells you to "switch to 2.5 Flash" in Gemini, you won't find that option. The closest label is plain Flash, which starts on Google AI Plus, and Google's help page doesn't tie it to a version number. For the bigger picture of the range, there's my guide to every Gemini model and what it's for, and the sibling page on Gemini 2.5 Pro covers the deeper-thinking model from the same generation.
See which model answered you
Because the app picks for you by default, it's worth knowing how to check. As of Nov 2026, on a computer:
- Go to gemini.google.com.
- At the top left, open the sidebar and click a chat.
- At the bottom of a response, click More.
- Select Show thinking steps.
- Scroll to the bottom of the thinking steps to see which model was used.
Switch model yourself
Manual switching needs a Google AI plan. If you have one:
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
- Go to gemini.google.com.
- At the bottom of the chat box, click the model that's currently selected.
- Click the model you want. Higher-tier models marked with a lock icon need an eligible plan upgrade.
- If you like, pick a thinking level: Low, Medium or High.
- Gemini keeps that model and thinking level for the rest of the conversation, including if you come back to it later.
To turn auto model selection on or off, go to Settings, then Usage limits, and use the switch at the bottom.
Thinking levels and your usage
Low gives fast, concise answers for straightforward questions. Medium balances speed and reasoning for everyday tasks. High lets the model work carefully through complex, multistep prompts. If you don't choose, Gemini defaults to Low.
Here's the trade-off that isn't obvious. Google says more advanced models and higher thinking levels use more of your usage. Gemini app limits refresh every 5 hours until you reach a weekly limit, and if you're on a Google AI plan and run out, your chat carries on with Flash-Lite. So picking the biggest model and High for every question burns through your allowance faster, often for no better answer on a simple job. My page on Gemini usage limits goes into how those limits work.
What I'd do
I'd stop planning anything new around 2.5 Flash. Google's own page says so, and it's not worth fighting the vendor's direction.
In the app, I'd leave Auto on and skip hunting for a 2.5 Flash setting, because it isn't there. When an answer matters, I'd check which model wrote it using Show thinking steps. For a team project on the API, I'd check Google's model page before choosing anything, since that note about access could change again.
Frequently asked questions
Is Gemini 2.5 Flash being shut down?
Google's developer page says the 2.5 models are not deprecated and will continue to be served until further notice through the API, with access limited to users who actively used them in the past. Only the preview version gemini-2.5-flash-preview-09-2025 is listed as shut down.
Can Gemini 2.5 Flash create images?
No. The developer page lists image generation and audio generation as not supported. It reads text, images, video and audio, but its output is text.
How much can Gemini 2.5 Flash read at once?
Google lists an input token limit of 1,048,576 and an output token limit of 65,536 for gemini-2.5-flash.
What is the knowledge cutoff for Gemini 2.5 Flash?
Google's model page gives the knowledge cutoff as January 2025, with the latest update in June 2025.
Which model does the free Gemini app use?
Without a Google AI plan you get Auto or Flash-Lite. With auto on, most prompts use Flash-Lite and harder ones may use Flash or Pro; with it off, all responses use Flash-Lite.
