Google's newest generally available model is Gemini 3.8 Flash, released on September 2, 2026. On September 30, 2026 Google announced Gemini 4 Argon, its next frontier model, and limited it at first to vetted cyber defenders in its Fairwind Program. The newest Pro model that developers can call is still Gemini 3.1 Pro, a preview from February 19, 2026. Gemini 3.5 Pro, announced at I/O on May 19, has not been released. This page lists each Gemini generation with dates, context windows, and the Google products each one runs in.
Gemini Model Timeline
| Model | Released | Context window | Notable change | Status |
|---|---|---|---|---|
| Gemini 1.0 Nano, Pro, Ultra | December 6, 2023 to February 8, 2024 | 32K | First Gemini generation. Nano ran on the Pixel 8 Pro | Discontinued |
| Gemini 1.5 Pro | February 15, 2024 | 1M, then 2M for developers from June 27, 2024 | 1M-token window in the production API | Discontinued |
| Gemini 1.5 Flash | May 14, 2024 | 1M | Faster, lower-cost tier | Discontinued |
| Gemini 2.0 Flash, Pro, Flash-Lite | January 30 to February 25, 2025 | 1,048,576 on Flash | Built around agents and speed | API shutdown June 1, 2026 |
| Gemini 2.5 Pro | March 25, 2025 | 1,048,576 | Reasoning model. January 2025 knowledge cutoff | API access limited to past users |
| Gemini 2.5 Flash, Flash-Lite | April 17 and June 17, 2025 | Not disclosed on current pages | Lower-cost 2.5 tiers | API access limited to past users |
| Gemini 3 Pro | November 18, 2025 | 1M | Launched in the Gemini app and AI Mode on day one | Preview |
| Gemini 3 Flash | December 17, 2025 | Not disclosed on current pages | Flash tier built on 3 Pro | Preview |
| Gemini 3.1 Pro | February 19, 2026 | 1,048,576 | Refinement of 3 Pro. Still the newest Pro on the API | Preview |
| Gemini 3.1 Flash-Lite | March 3, 2026 | Not disclosed on current pages | High-volume, low-latency tier | Shutdown set for May 7, 2027 |
| Gemini 3.5 Flash | May 19, 2026 | 1,048,576 | Became the AI Mode default at I/O | Stable |
| Gemini 3.5 Pro | Announced May 19, 2026 | Not released | Promised for the month after I/O | Not released |
| Gemini 3.6 Flash, 3.5 Flash-Lite | July 21, 2026 | 1,048,576 | Coding and multimodal gains | Stable |
| Gemini 3.7 Flash | August 13, 2026 | 1,048,576 | Coding and agents. Google's pick for efficiency-first work | Stable |
| Gemini 3.8 Flash | September 2, 2026 | 1,048,576 | Long-running engineering and agent tasks | Latest stable model |
| Gemini 4 Argon | September 30, 2026 | Input not disclosed. Output up to 1M tokens | Google's new frontier model | Limited preview |
Since May 2026 every Gemini release that developers could use has been a Flash model: 3.5, 3.6, 3.7, and 3.8 Flash in under four months. Google's API changelog lists no Gemini 3.2 release, so that name is not in this table. Output on the 3.x models we checked is capped at 65,536 tokens. The three newest Flash models share an introductory price of $0.75 input and $3.75 output per million tokens through December 31, 2026, doubling on January 1, 2027.
Which Surface Runs Which Model
| Surface | Model as of October 1, 2026 |
|---|---|
| AI Mode in Google Search | Gemini 3.5 Flash has been the default for all users since May 19, 2026. Google AI Pro and Ultra subscribers get Gemini 3.8 Flash |
| AI Overviews | Merged with AI Mode into one AI Search experience at I/O 2026. Flash-tier models typically back it |
| Gemini app | Gemini 3.8 Flash for AI Pro and Ultra subscribers. Coverage from eesel and others says the free plan does not include it |
| Gemini Spark agent | Gemini 3.7 Flash rolled out on August 13 for AI Pro and Ultra, according to 9to5Google |
| Gemini API, AI Studio, Antigravity | Gemini 3.8 Flash. Google tells new projects to use 3.8 Flash or 3.5 Flash-Lite |
| Gemini in Google Sheets | Gemini 3.8 Flash |
| Siri AI on iOS 27 | Apple Foundation Models built in collaboration with Google's Gemini |
| Fairwind Program | Gemini 4 Argon and Gemini 3.8 Flash Cyber, for trusted defenders only |
See AI Mode's switch to Gemini 3.5 Flash and Siri AI on iOS 27.
What Is Next
Gemini 4 Argon for everyone else. Google's announcement says Argon will reach developers, enterprises, and consumers as soon as possible, starting with paid API customers and Google AI Ultra subscribers. It gives no dates and no public model ID. The introductory price is $2 input and $10 output per million tokens, rising to $4 and $20.
Gemini 3.5 Pro. Codersera's tracker cites a Bloomberg report from July 16 that Google was taking more time to improve the model, mainly on coding, and a Google statement from July 21 that it was testing with partners. The Argon announcement does not mention 3.5 Pro. Whether it still ships is unknown.
Price change. The Flash introductory rates end on December 31, 2026.
Brand Visibility Implications
- Google shipped four Flash models in under four months. Each one that reaches AI Mode or the Gemini app can change which brands and sources appear, with no announcement inside Search.
- Paid and free users see different models. AI Pro and Ultra subscribers get 3.8 Flash in AI Mode and the Gemini app. Track both where you can.
- Google does not list a knowledge cutoff on its 3.x Flash model pages. The last one we could confirm is January 2025 for Gemini 2.5 Pro. AI Mode and AI Overviews ground answers in Search, so retrieval carries much of the weight there, but the model's own recall still shapes which brands it thinks to look for.
- Older models persist. Gemini 2.5 remains callable for existing API users, and products built on it still answer from early 2025 knowledge.
See how training cutoffs change brand recall.
Methodology
This page is built from vendor documentation and dated reporting, not from Presenc AI measurements. Primary sources are used wherever one exists: OpenAI's model pages and deprecations page, Google's Gemini API changelog and deprecations page, xAI's release notes and model list, DeepSeek's API changelog and pricing page, and MiniMax's pricing page and Hugging Face model cards. Older release dates that vendors no longer document come from the relevant Wikipedia articles. Where a figure comes only from a secondary tracker or a single outlet, the page names it. "Not disclosed" means the vendor material we read does not state the figure. Status as of October 1, 2026.
How Presenc AI Helps
Presenc AI tracks how Gemini, AI Mode, and AI Overviews describe and cite your brand over time. When Google swaps the model behind a surface, you can see whether your visibility moved with it.