Gemini (app)
Last reviewed
Apr 30, 2026
Sources
25 citations
Review status
Source-backed
Revision
v1 ยท 3,967 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Apr 30, 2026
Sources
25 citations
Review status
Source-backed
Revision
v1 ยท 3,967 words
Add missing citations, update stale details, or suggest a clearer explanation.
The Gemini app is Google's consumer-facing AI chat application, available on the web at gemini.google.com, as a native Android app, and as a native iOS app. It is powered by the Gemini family of large language models built by Google DeepMind, and competes directly with ChatGPT, Microsoft Copilot, and Anthropic's Claude in the consumer assistant market.
The product was originally launched as Bard on March 21, 2023, after being announced by Google CEO Sundar Pichai on February 6, 2023. On February 8, 2024, Google rebranded Bard to Gemini, simultaneously launching a paid tier called Gemini Advanced powered by the Gemini Ultra 1.0 model. Since then the app has been the main vehicle through which Google ships consumer AI features, including multimodal input, long-context reasoning, image generation via Imagen, video generation via Veo, the conversational Live mode, the agentic Deep Research feature, and custom assistants called Gems.
By late 2025 the Gemini app reported more than 650 million monthly active users, making it one of the largest consumer AI products in the world, second to ChatGPT in raw scale but tightly integrated with Google's much larger Search, Workspace, and Android footprints.
Google's path to a public chatbot began long before ChatGPT. In 2020 it published Meena, an early dialogue model, then announced LaMDA (Language Model for Dialogue Applications) at Google I/O 2021. The full LaMDA paper, by Thoppilan et al., appeared in early 2022. LaMDA was a 137-billion-parameter dialogue model fine-tuned on conversational data, with a strong emphasis on safety and groundedness. Internally Google saw it as a research artifact rather than a product, and was reluctant to open it up to the public for reasons mixing reputational caution with revenue concerns about cannibalising Search.
On November 30, 2022, OpenAI released ChatGPT. Within two months it had crossed 100 million users. CNBC reported on January 31, 2023 that Google had declared an internal "code red" and was testing several internal chatbots, one of them called Apprentice Bard, built on LaMDA. Microsoft's roughly $10 billion expansion of its OpenAI partnership, announced on January 23, 2023, intensified the pressure by tying ChatGPT-class models to Bing and Office.
On February 6, 2023, Sundar Pichai published a blog post titled "An important next step on our AI journey" announcing Bard, an experimental conversational AI service powered by a lightweight version of LaMDA. The next day, February 8, 2023, Google held a livestream in Paris meant to showcase Bard alongside a refreshed Search experience. Bard incorrectly claimed that the James Webb Space Telescope had taken "the very first pictures of a planet outside of our own solar system," a credit that actually belongs to the European Southern Observatory's Very Large Telescope from 2004. The error spread on social media, and Alphabet's stock dropped about 8% the following day, wiping out roughly $100 billion in market capitalisation. The episode set the tone for early coverage of Bard as a hurried response to ChatGPT.
Bard opened to early access via waitlist on March 21, 2023, in the United States and the United Kingdom. The initial product was a single-turn-friendly chat interface on bard.google.com. It used a lightweight LaMDA derivative, supported only English, and shipped with conspicuous "experiment" labelling. Reviews were mixed: faster than ChatGPT for some queries, weaker on reasoning, and visibly hedged.
At Google I/O on May 10, 2023, Pichai announced that Bard had moved off LaMDA and onto PaLM 2, a more capable general-purpose model. The waitlist was dropped and Bard rolled out to more than 180 countries and territories. Japanese and Korean joined English, with 40 more languages promised.
On July 13, 2023, Bard expanded into the European Union and Brazil after months of regulator engagement, particularly with the Irish Data Protection Commission. The same update added image input via Google Lens, audio responses, pinned chat history, and tone controls (simple, long, short, professional, casual). It now supported 40+ languages.
On December 6, 2023, Google announced Gemini 1.0, its first natively multimodal model family, in three sizes: Ultra, Pro, and Nano. The same day a specifically tuned version of Gemini Pro began powering Bard in English in over 170 countries. Google also previewed "Bard Advanced" with Gemini Ultra for early 2024.
On February 8, 2024, Google rebranded Bard to Gemini across all platforms, launched a new dedicated Gemini app on Android, made Gemini available through the Google app on iOS, and introduced Gemini Advanced as a paid tier on a new Google One AI Premium plan priced at $19.99 per month in the United States. Gemini Advanced was the first place consumers could use Gemini Ultra 1.0, which Google described as the first model to outperform human experts on the MMLU benchmark across 57 subjects.
The rebrand consolidated several earlier brands. Bard merged into Gemini on the consumer side, and Duet AI for Workspace and Duet AI for Google Cloud were renamed to Gemini for Workspace and Gemini for Google Cloud. The new app shipped in 40 languages on the web, with Android availability rolling out first to the United States.
| Date | Milestone |
|---|---|
| Feb 6, 2023 | Bard announced by Sundar Pichai, powered by a lightweight LaMDA. |
| Feb 8, 2023 | Paris livestream demo includes James Webb factual error. Alphabet stock drops about 8%. |
| Mar 21, 2023 | Bard early access opens via waitlist in US and UK. |
| May 10, 2023 | Google I/O: Bard moves to PaLM 2; expanded to 180+ countries; waitlist removed. |
| Jul 13, 2023 | Image input via Google Lens; EU and Brazil launch; 40+ languages; audio responses. |
| Dec 6, 2023 | Gemini 1.0 announced. Gemini Pro begins powering Bard in 170+ countries in English. |
| Feb 8, 2024 | Bard rebranded to Gemini. Gemini Advanced launches with Ultra 1.0 on Google One AI Premium ($19.99/mo). Native Android app ships. |
| Feb 22, 2024 | Image generation of people paused after historical inaccuracy controversy. |
| May 14, 2024 | Google I/O: Gemini 1.5 Pro with 1M-token context becomes available in Gemini Advanced; Project Astra and Gemini Live previewed. |
| Aug 13, 2024 | Gemini Live (voice mode) launches; Gems custom assistants become available; Imagen 3 lands. |
| Nov 14, 2024 | Native iOS Gemini app launches worldwide on the App Store. |
| Dec 11, 2024 | Gemini 2.0 Flash launches in the app; Deep Research feature debuts; Project Mariner browser agent previewed. |
| Feb 5, 2025 | Gemini 2.0 Flash reaches general availability; 2.0 Pro Experimental and 2.0 Flash-Lite arrive. |
| Mar 25, 2025 | Gemini 2.5 Pro Experimental, Google's first "thinking" reasoning model. |
| May 2025 | Google I/O: Gemini 2.5 family expands; Veo 3 with synchronised audio; new AI subscription plans (AI Pro, AI Ultra). |
| Aug 2025 | Gemini Live adds visual help and deeper Google app connections. |
| Oct 29, 2025 | Alphabet Q3 2025 earnings: Gemini app reports more than 650 million monthly active users. |
| Nov 18, 2025 | Gemini 3 Pro launches across the Gemini app, AI Studio, and Vertex AI. |
| Dec 17, 2025 | Gemini 3 Flash launches with Pro-level performance. |
| Feb 4, 2026 | Reports indicate the Gemini app has surpassed 750 million monthly active users. |
| Feb 19, 2026 | Gemini 3.1 Pro preview rolls out across consumer and developer products. |
From the Gemini 1.0 launch onward the app has accepted text, images, audio, and uploaded files. Gemini 1.5 Pro added native video understanding in May 2024 along with a 1 million token context window, enough to ingest roughly 1,500 pages of text, an hour of video, or codebases over 30,000 lines in a single prompt. A 2 million token window was made available to developers in private preview. Gemini 3 Pro currently supports up to 1,048,576 input tokens with up to 65,536 output tokens.
The app generates images via Google's Imagen line of diffusion models. Imagen 2 launched alongside the rebrand in February 2024 and was paused later that month after the historical-figure controversy. Imagen 3 returned image generation in August 2024 with stronger guardrails, and Imagen 4 followed in 2025 with up to 2K resolution. Video generation is powered by Google DeepMind's Veo model: Veo 2 became available to advanced users in April 2025, and Veo 3 followed in May 2025 with synchronised audio (dialogue, sound effects, ambient noise) generated alongside visuals. Veo 3.1 added more controllability in October 2025. Video generation is gated to higher-priced subscription tiers.
Gemini Live, previewed at I/O on May 14, 2024 and launched on August 13, 2024, is a real-time voice conversation mode. It supports interruption mid-response, switching topics, and a selection of natural-sounding voices. Visual Live, which lets the model see through the phone camera or share a screen, rolled out in 2025.
Deep Research, launched on December 11, 2024, takes a complex query and produces a multi-step research plan that the user approves before execution. The agent then browses dozens to hundreds of web pages over several minutes, with newer versions also able to draw on Gmail, Drive, and Chat content. The output is a multi-page report that can be exported to Google Docs. Deep Research Max, a more autonomous variant, arrived in 2026.
Gems, launched on August 13, 2024, are custom mini-assistants that users define with a name and instructions. They are roughly equivalent to OpenAI's GPTs. Premade Gems include a learning coach, brainstorming partner, and coding partner. Gemini Extensions for Workspace let the app pull from Gmail, Drive, Docs, Calendar, and Maps with the user's permission. Inside Workspace itself, Gemini powers "Help me write" in Gmail and Docs, "Help me organise" in Sheets, "Help me visualise" in Slides, and meeting summaries in Meet. The side-panel Gemini in Docs, Sheets, Slides, and Drive rolled out in June 2024.
Project Mariner, previewed in December 2024, is an experimental Chrome-based browsing agent that reads the page and performs tasks on the user's behalf. It scored 83.5% on the WebVoyager benchmark at launch. Mariner began rolling into the Gemini app for AI Ultra subscribers in 2025. Jules, also previewed in December 2024, is a code-focused agent. The app supports code generation, code execution in a sandboxed Python environment (the equivalent of OpenAI's Code Interpreter), and structured data analysis on uploaded CSVs. Search grounding is available across most query types, with citations linking back to the underlying web pages.
Google's consumer subscription structure has evolved considerably since the Gemini Advanced launch. As of early 2026 the lineup looks like this:
| Plan | Monthly price (US) | Models | Notable features |
|---|---|---|---|
| Free | $0 | Gemini Flash and limited Pro access | Basic chat, image input, lightweight Imagen, modest Live, limited Deep Research |
| Google AI Plus | $7.99 | Pro models with raised caps | Higher limits than free; entry tier introduced 2025 |
| Google AI Pro (formerly Google One AI Premium) | $19.99 | Gemini 3 Pro / 2.5 Pro thinking, 1M context | 100 prompts/day with Pro, 20 Deep Research reports/day, Veo 2 access, 2 TB Google One storage, Workspace integration |
| Google AI Ultra | $249.99 | Highest caps, Deep Think, Veo 3, Nano Banana 2 image | 500 prompts/day, 200 Deep Research reports/day, 1,000 image generations/day, Project Mariner, 30 TB storage, YouTube Premium |
| Gemini for Workspace (Business / Enterprise) | varies | Pro models with admin controls | Domain controls, data residency, audit logs, Workspace integration |
| Gemini for Education | bundled with edu plans | Pro models | School-specific safeguards, NotebookLM access |
The Google One AI Premium plan was renamed to Google AI Pro in 2025, and the new top-end Google AI Ultra plan was introduced at I/O 2025 alongside Veo 3 and Deep Think. New subscribers typically receive promotional discounts of 50% off for the first 2 to 12 months depending on the tier.
The Gemini app sits in a crowded consumer AI market. The main rivals as of early 2026 are:
| Product | Maker | Launched | Free tier | Paid entry price | Multimodal in/out | Live voice | Image gen | Video gen | Agent / browsing |
|---|---|---|---|---|---|---|---|---|---|
| Gemini app | Mar 21, 2023 (as Bard) | Yes | $7.99 (AI Plus) / $19.99 (AI Pro) | Text, image, audio, video, files | Yes (Live) | Yes (Imagen) | Yes (Veo) | Yes (Deep Research, Mariner) | |
| ChatGPT | OpenAI | Nov 30, 2022 | Yes | $20 (Plus) / $200 (Pro) | Text, image, audio, video | Yes (Advanced Voice) | Yes (gpt-image / DALL-E) | Yes (Sora) | Yes (Operator, Agent) |
| Claude | Anthropic | Mar 2023 (Claude 1) | Yes | $20 (Pro) / $200 (Max) | Text, image, files | No native voice | No native | No native | Yes (Computer Use, Skills) |
| Microsoft Copilot | Microsoft | Feb 2023 (as Bing Chat) | Yes | $20 (Copilot Pro) | Text, image, voice | Yes | Yes (DALL-E) | Limited | Yes (Copilot Vision, Actions) |
| Perplexity | Perplexity AI | Dec 2022 | Yes | $20 (Pro) / $200 (Max) | Text, image, files | Yes | Yes | Limited | Yes (Comet) |
| Meta AI | Meta | Sep 2023 | Yes | n/a (free) | Text, image | Yes | Yes (Imagine) | Yes (Movie Gen) | Limited |
| DeepSeek | DeepSeek | 2024 | Yes | n/a (free) | Text, image | No native | Limited | No | Limited |
Gemini's distinct advantages are its tight Google ecosystem integration (Workspace, Search, Android, Pixel), the longest widely available context window, free-tier video generation in some markets, and the Deep Research agent. ChatGPT remains ahead on raw user count and on developer mindshare; Claude remains ahead on long-form writing and code review; Copilot wins on Microsoft 365 integration. The competitive picture has been a moving target, with each model release temporarily putting one product ahead on benchmarks before the others catch up.
The most damaging episode in the app's short history began in late February 2024, weeks after the Gemini rebrand. On February 21 and 22, users on X posted screenshots of Gemini producing historically implausible images. Asking for "a 1943 German soldier" returned nonwhite people in Nazi uniforms. Asking for "the Founding Fathers of the United States" returned an Indigenous-American man, a Black man, and an Asian man in colonial dress. Asking for a Pope returned a Black woman. Some prompts asking for white people were refused outright, while equivalent prompts for other races returned images.
The behavior was the result of an over-aggressive system prompt that injected diversity-related instructions into the user's request before sending it to Imagen 2. The system was attempting to counteract well-known biases in image models that under-represent women and people of color, but it did so without exceptions for cases where historical accuracy or specificity mattered.
On February 22, 2024, Google paused image generation of people in Gemini, posting on X that it would "pause the image generation of people and re-release an improved version soon." The next day, Prabhakar Raghavan, head of Google Search, published a blog post calling the outputs "embarrassing and wrong." Pichai sent an internal memo, later reported by Semafor, calling the issue "completely unacceptable."
Image generation of people stayed paused for roughly six months. It returned in late August 2024 with the upgraded Imagen 3 model and a different mix of guardrails. The episode reinforced two perceptions that already hung over Gemini: that Google was over-correcting on safety and bias, and that its launches were rushed. It was repeatedly cited by critics through 2024 and 2025 in conversations about whether AI products were ideologically tilted.
Hallucinations. Like all large language models, Gemini fabricates facts and citations, especially in the free tier. Reviews of the Bard launch in March 2023 noted frequent confident errors, and even later 2.x and 3.x models continue to hallucinate at a measurable rate on long-tail factual questions and recent events.
Search substitution. AI Overviews, which use a Gemini variant in Search, drew complaints from publishers about reduced click-through rates. A set of viral examples in May 2024 showed AI Overviews suggesting users put glue on pizza and eat rocks; Google narrowed the feature's scope afterwards. Antitrust pressure on Google's Search business adds further complications, as the US Department of Justice's pending remedies in the Search antitrust case have at times included proposals that touch on AI distribution.
Privacy. Free users' conversations are by default available for product improvement and human review, though Google introduced clearer opt-out paths and shorter default retention windows in 2024 and 2025. EU rollouts have been gated by the Irish Data Protection Commission and the broader EU AI Act framework.
Safety incidents. In November 2024 a CBS News report documented a Gemini response that told a student researching elder care to die. Google described it as a policy violation and patched the model. In 2026 a wrongful-death lawsuit was filed alleging that Gemini's responses contributed to a user's suicide; the case remains pending.
Adoption gap. Even at 650 million MAU in Q3 2025 and a reported 750 million by early 2026, Gemini still trailed ChatGPT in raw consumer adoption, according to data cited by PYMNTS, TechCrunch, and similarweb. Coverage frames this as a structural disadvantage despite Google's distribution power.
Gemini is no longer just an app; it is increasingly the surface across which Google's AI is delivered.
Pixel phones. The Pixel 8 Pro was the first phone engineered for Gemini Nano, the on-device model in the Gemini family, with features like Summarize in Recorder and Smart Reply in Gboard. The Pixel 9 series upgraded to multimodal Gemini Nano, powering Pixel Screenshots, Talkback image descriptions, call summaries, and Pixel Weather. From the Pixel 9 onwards, Gemini is the default voice assistant rather than the legacy Google Assistant. The Pixel 10 lineup brought further on-device capabilities.
Android. Beyond Pixel, Gemini is rolling out as the default assistant across Android phones from Samsung, Motorola, OnePlus, and others, with Google Assistant being formally retired through 2026. Samsung Galaxy S25 and Z Fold/Flip 7 devices ship with Gemini deeply integrated.
iPhone, Chrome, Wear OS, Nest. The native iOS Gemini app launched on November 14, 2024, after months of being available only through the Google iOS app. It supports text in 35 languages and Live in 12. Gemini integration in Chrome's address bar lets users invoke the assistant while browsing. Gemini began landing on Pixel Watches in 2025 and is gradually replacing the legacy Assistant on Nest speakers and displays.
Search and Workspace. AI Overviews, AI Mode, and the underlying generative summaries in Google Search are powered by Gemini-derived models, reaching over a billion users a month, vastly larger than the standalone app. Gemini for Workspace embeds the assistant in Gmail, Docs, Sheets, Slides, Drive, Meet, and Chat for paying customers, with admin controls for Business and Enterprise plans.
Google has gradually disclosed more numbers over time. The reported milestones are:
| Date | Reported Gemini app MAU |
|---|---|
| Oct 2024 | ~90 million |
| Mar 2025 | ~350 million (court documents) |
| Q3 2025 (Oct 29, 2025 earnings call) | 650+ million |
| Q4 2025 / early 2026 (TechCrunch) | 750+ million |
These numbers do not include users who interact with Gemini-powered features inside Search, Workspace, or Android. Including those surfaces, Google has claimed Gemini reaches well over a billion users monthly, though those figures are not always strictly comparable to ChatGPT's MAU.
The Gemini app is the consumer-facing surface of Google's broader AI strategy, sitting alongside the search-and-ads front (AI Overviews, AI Mode) and the enterprise front (Gemini API, Vertex AI, Gemini for Workspace, Gemini for Cloud). Google's distribution gives it reach no rival can match, but also means any regression makes headlines. The shift from the cautious Bard of March 2023 to a 750-million-MAU multimodal assistant in early 2026 happened in under three years, a pace few consumer products have matched.