Gemini

There are two things to consider: where you are located and whether you need an assistant inside Google services.

Who it is for. For those with Gmail email, Drive files, and an Android phone: Gemini is already there and reads your emails and documents on request. And for those who want a strong model for free: Google's Pro model is available without a subscription; you pay for volume (Google Help, 05.09.2026).

Who else. For those who work with large files: for $19.99 per month, the context window grows to a million tokens, several books in one conversation. It is also enough for images and short videos: 150 million images are created in Gemini every day (blog.google, August 2026).

Who it is not for. For those in Russia without a permanent foreign IP address: Russia is not on Google's country list, and on August 13-14, 2026, some users lost access even from foreign addresses. And for those who need a coding agent in the terminal: Google shut down its Gemini CLI for subscribers on 18.06.2026, and coding moved to Antigravity.

A good fit: section 3 will take you from a foreign address to your first prompt. Choosing between it and ChatGPT: see the decision point in section 7.

What vibecoding isStarting from scratch: how to set tasks for agents with no coding experience

Three layers: a chat with three models, tools inside Google services, and agents that work without you.

Chat and models. The selector has three models: the fast Flash-lite, the main Flash, and the thoughtful Pro, and all three are available without a subscription. You choose response depth separately: regular, long, and Deep Think, which thinks for several minutes and is only available in Ultra. The models have version numbers in the API, and the smartest one now is Gemini 3.8 Flash (02.09.2026), while Pro has been on version 3.1 since February.

Google tools. Connected apps give Gemini access to Gmail, Drive, Maps, and YouTube: "find the contract sent in March" works without uploading files. Canvas opens a document or code window next to the chat, Deep Research independently assembles a report from dozens of sources, and Notebook keeps a project with sources, up to 600 in one.

Images, video, voice. Nano Banana 2 creates images, while Nano Banana Pro edits them. Gemini Omni Flash makes videos in the app: 3-10 second clips since 19.05.2026, while the previous Veo 3.1 remains in the Flow studio and in the API. Live voice mode is only available in the mobile app, not on the web.

Agents. Gemini Spark, a personal agent that handles tasks on its own, becomes available with an AI Pro subscription; the feature list includes the Daily Brief morning summary. The feature list also includes scheduled actions. Gemini Agent and calls to local businesses are currently available only in the US: the agent in Ultra, automated website browsing in Chrome with AI Pro.

What was actually done. We did not run Gemini through our own tasks; the figures come from Google Help and third-party tests. Users praise response speed and the million-token context window, which iphones.ru calculated as 750 thousand words, more than twice ChatGPT's amount (26.01.2026). RuStore reviews from August 2026 criticize freezing chats and refusals to generate under safety rules, while praising videos and images.

You can use it from Russia, but not directly: first the address, then the account.

Step 1. Change your Russian IP address to a foreign one: US, Dutch, or any other from Google's country list. The service is officially unavailable in Russia, so without this the page will say "not supported in this country."

Step 2. Create a Google account set to the same country. Google checks not only the address, but also the account country and payment profile: on August 13-14, 2026, some Russians with foreign addresses lost access, and it returned for some but not others (kod.ru, RuStore, August-September 2026). There is no guarantee.

Step 3. Open gemini.google.com and ask your first question: no registration beyond a Google account is required. Write in Russian and it will answer in Russian: Russian is on the official language list.

Apps. The official app is available only in foreign storefronts: the US or Kazakhstan App Store and Google Play with a non-Russian account region. In the Russian App Store, searching for "gemini" returns clones called "Gemi AI" and "Genimi", while RuStore's "Gemini" listing was published by third-party developer GoodApp and has not been updated since 12.03.2025. It is not a Google app, even though Alice in Yandex calls it official.

Computer. The macOS app launched on 15.04.2026: it requires Apple Silicon and macOS 15 and opens with Option + Space. There is no Windows app; use the web or Gemini in Chrome.

Steps verified on 05.09.2026 against Google Help (countries and languages), the App Store and RuStore storefronts, kod.ru, and practitioner reviews

Three signals diverge: record audience, solid models, but not the top ones.

Audience. One billion monthly users as of 11.08.2026, Google's fourteenth product to reach that number and the fastest-growing in its history (blog.google, TechCrunch). It had 900 million at Google I/O in May.

LMArena. Google's best model in the text arena, Gemini 3.8 Flash, ranks eighth with 1494 points and a "preliminary" label (snapshot 03.09.2026); the first seven spots are held by Anthropic and Meta. In the web development arena, the best Gemini is in 16th place. Google's smartest model is now called Flash, not Pro: Pro 3.1 is fifteenth in the text arena.

Artificial Analysis index. Gemini 3.8 Flash scores 59 points versus 66 for the leader, Claude Fable 5.1 (05.09.2026). Gemini 3.1 Pro scores 48 there, below Google's own Flash models.

Demand in Russia. "Gemini" is searched 696 thousand times a month, "gemini" 259 thousand, "dzhemini" 88 thousand, and these figures include noise from the zodiac sign and the film; the specific query "gemini in russia" accounts for 23 thousand (Wordstat, 05.09.2026). Gemini's place among the assistants in our catalog is shown below the card title, while the full tool ranking is on the main page.

Models are free; you pay for limits, context, and cloud storage.

Free entry. All three models, including Pro, are available without a subscription, with 32 thousand tokens of context and a standard limit that resets every 5 hours. Deep Research and media generation use up the limit faster.

Paid tiers. AI Plus for $4.99 doubles the limit, AI Pro for $19.99 gives a limit four times higher, one million tokens of context, and the Spark agent, while AI Ultra for $99.99 and $199.99 adds Deep Think. Prices are per month; Google has no annual prices on the page (05.09.2026).

Paying from Russia. Russian cards do not work: Google does not accept Mir at all, and has declined Russian Visa and Mastercard since 2022. With the Russian locale, the storefront displays rubles, 1 790 ₽ for AI Pro, but a purchase with a Russian account does not go through. The usual options are a foreign card, an app store with a foreign region, or an intermediary, where AI Pro costs 1 500-2 200 ₽. Details are in the guide how to pay for AI services from Russia.

For developers. The API charges $0.75 per million input tokens and $3.75 per million output tokens for Gemini 3.8 Flash, but that is the price until 31.12.2026; it doubles in January. Pro 3.1 costs $2 and $12, and has no free API tier.

Three questions: what Google does with chats, what the agent can do on its own, and how much to trust its answers.

Chats are used for training. History is on by default, and Google explicitly says it uses it "to improve services, including training generative models." Some chats are read by human reviewers, and those copies, disconnected from the account, are stored for up to three years and are not deleted with the history. Turn off history and chats still remain for 72 hours (Privacy Hub, updated 10.08.2026).

The agent makes mistakes for you. Google itself warns that Gemini "may make mistakes, for example, make a purchase or unexpectedly share your data with a third party," and asks you to monitor its actions in the browser. Google does not review third-party connected apps.

Answers need fact-checking. Practitioners complained in August 2026 about chats freezing and going blank, and refusals to generate images under safety rules. Google itself places the line "Gemini is AI and can make mistakes" below answers, so verify the numbers.

What else to know. In July 2026, publishers filed a US lawsuit alleging that Gemini was trained on their books; there has been no ruling on the merits. The service's legal entity is Google LLC, and Google Ireland in the EEA and Switzerland.

The choice is not "who is smarter," but "which ecosystem you live in."

Different environment. ChatGPT wins on tool coverage, memory across chats, and a desktop app with the Codex agent. Gemini wins because it is already built into Gmail, Drive, and Android, with one million tokens of context and a free Pro model.

Different entry price. ChatGPT has a free plan and a $20 Plus subscription, while Gemini's first tier costs $4.99. Neither is available directly from Russia, and neither can be paid for with a Russian card.

Who else is nearby. If you need an assistant that opens from Russia without setup, look at Alice AI and DeepSeek: they are the ones listed alongside clones when searching for "gemini" in the Russian App Store.

In three years, Bard became an assistant for one billion people, while the smartest model moved from the Pro line to the Flash line.