![Gemini Drops: 11 Features, 7 Reach Japan [2026]](https://shattered.io/wp-content/uploads/2026/11/gemini-drops-october-2026-11-features-argon-antigravity-skills-2026-gen.webp)
Google bundled 11 features into its Gemini Drops October 2026 release, and the headline item is a coding agent that now lives inside the Gemini app on macOS. According to jetstream.blog, Google’s @GeminiApp account announced the batch on October 30, 2026, though other outlets date the announcement to October 29. The list spans a new model, two desktop apps, a prompt-saving system, file editing, music generation and avatar sharing.
The drop is less a single launch than a stack of rollouts that Google folded into one monthly package. Several pieces shipped weeks ago, including the Windows app on September 10 and the model on October 13. That timing matters for anyone reading the headline number. Eleven features sounds like a flood, yet only a handful went live this week. The rest repackage earlier releases under the Drops banner.
This analysis walks through each feature, sorts what is new from what is recycled, maps the regional gaps, and weighs what the package means for the AI assistant race. It also builds on our earlier look at how Chrome’s Gemini autofill reads Gmail across three data types, since both stories point at the same strategy of pulling personal data deeper into Gemini. Everything below comes from the reports we could verify. Where coverage is thin, we say so.
Don't miss new tech stories on Google
Add Shattered once in the Google app and our stories show up in your news suggestions.
What Gemini Drops October 2026 Actually Contains #
Gemini Drops is Google’s recurring feature roundup for the Gemini app. The October 2026 edition lists 11 items: Gemini 4 Argon, a macOS Gemini app with Google Antigravity, a Windows Gemini app, Skills, deeper Google Maps support for Personal Intelligence, Gemini Notebook integration, expanded app connectors, file creation and editing, a student tool called “Step inside with immersive explorations,” Lyria 3.5 for music, and shareable avatars.
Jetstream.blog published its breakdown on October 31, 2026, and a second outlet, Creators Toolbox, ran its own version on October 30. Both agree on the core list. A Substack newsletter, AI Matter In, counted the same 11 features and noted that 7 of them reach Japan, which gives us a useful proxy for how unevenly the package rolls out worldwide.
Reports also differ on who gets what. One summary says the features reach all Gemini users, including people without a paid plan. Jetstream’s detail contradicts that for at least the headline model, which sits behind Google AI Pro and Google AI Ultra. We use the more conservative reading throughout. Treat “available to everyone” as unproven until Google publishes a per-feature entitlement table.
The drop also shows how Google now ships. Instead of one keynote moment, it runs a steady cadence of small releases and then bundles them monthly. That keeps Gemini in the news every four weeks and lets Google count older launches toward a bigger number. It works as marketing, and it complicates anyone trying to measure real velocity.
The 11 Features at a Glance #
The table below condenses each feature, the date attached to it in the reports, and the main constraint. Dates come from jetstream.blog unless noted. Where a source gave no date, we leave the cell as “not stated” rather than guess.
| # | Feature | Reported date | Main constraint |
|---|---|---|---|
| 1 | Gemini 4 Argon | Oct. 13, 2026 | Google AI Pro and Ultra |
| 2 | macOS app with Antigravity | Not stated | English only, age 18+ |
| 3 | Windows Gemini app | Sept. 10, 2026 | Windows 10 or later |
| 4 | Skills | Oct. 13, 2026 | Gems migrate from Nov. 17 |
| 5 | Maps in Personal Intelligence | Gradual rollout | Staged by region |
| 6 | Gemini Notebook integration | Not stated | Off by default in the EEA |
| 7 | Expanded connectors | Mid-October rebrand | Mostly US English |
| 8 | File creation and editing | Oct. 29, 2026 | Most formats web only |
| 9 | Immersive explorations | Oct. 28, 2026 | Students hub, 6 languages |
| 10 | Lyria 3.5 | Announced Sept. 4, 2026 | Global |
| 11 | Shared avatars | Rolling out | English, outside EEA, UK, Switzerland |
Read the dates column and the pattern jumps out. Two of the 11 items shipped or were announced in September, two more arrived on October 13, and only the file editing and immersive explorations arrived in the last few days. The October label reflects Google’s packaging, not a single release day.
Gemini 4 Argon: The Model Behind the Drop #
Gemini 4 Argon is the first model in Google’s new Gemini 4 family, and jetstream.blog reports it became generally available worldwide on October 13, 2026. It runs in the Gemini app and in AI Mode in Search, and Google lists it as a selectable option next to Gemini 3.1 Pro rather than a replacement. Access requires a Google AI Pro or Google AI Ultra subscription.
Google’s pitch centers on professional work. Per jetstream’s summary of the announcement, the company says Argon performs at a top tier in programming, finance, law and cybersecurity defense. Those are four of the most lucrative verticals for enterprise AI spending, and the choice of examples signals where Google wants to compete.
The coverage we found publishes no benchmark scores. AI Matter In states plainly that the text offers no test results for Argon. That gap deserves attention. A model claim with no numbers attached is a marketing statement, and readers should hold it that way until independent evaluators report. Our own coverage of AI code security showed how far apart vendor claims and third-party testing can land, with Claude Opus 5.5 and GPT-6-Sol tying at 73% in that study.
Why keeping Gemini 3.1 Pro matters
Leaving Gemini 3.1 Pro in the picker tells us something. Google did not retire the older model on day one, which suggests either capacity limits on Argon or a wish to let users compare results. Teams that built prompts around 3.1 Pro can keep running them while they test the newer model. For developers, that is the safer migration path, and it costs Google some of the clean narrative that a hard cutover would give.
Antigravity on Mac: Chat Becomes a Coding Agent #
The feature drawing the most attention is Google Antigravity inside the macOS Gemini app. According to Creators Toolbox, a button at the top of the app launches the agent. You approve its plan, return to chat, and the agent builds and tests while you do something else. Jetstream describes the use cases as building and debugging software, prototyping interactive 3D games and automating workflows.
The restrictions are real. The Mac build supports English only, requires users to be 18 or older, and works only where Antigravity itself is supported. Creators Toolbox also lists exclusions for the European Economic Area, Switzerland and the United Kingdom. It covers personal accounts and eligible work accounts. Japan does not get it at launch, per jetstream.
The approval step is the design choice worth studying. Google puts a human checkpoint between the plan and the execution, which mirrors a pattern spreading across the industry. We covered a close cousin in Claude Duet, which uses four approval gates and two plans. Both products bet that users want an agent that moves fast but asks before it acts. Google’s version uses one gate on the plan, so it trades some caution for speed.
Putting a coding agent inside a consumer chat app also changes who builds software. A person who never opened a terminal can now describe a small game or an automation and watch it get built and tested locally. Whether the output holds up is a separate question, and the security results cited above suggest it often needs review.
Windows App and the Alt + Space Shortcut #
The Windows Gemini app launched on September 10, 2026, and Google folded it into the October package. Users summon it with Alt + Space, which lets them check a fact or brainstorm without leaving the window they are working in. It requires Windows 10 or later and works in every country and language where web Gemini runs, Japan included.
Jetstream lists three capabilities in the app. Gemini Spark handles task delegation. Nano Banana generates images. Gemini Omni generates and edits video. Google gave no usage limits for these in the sources we reviewed, so we cannot say how many generations a free account gets versus a paid one.
The shortcut matters more than the feature list. A global hotkey turns an assistant from a destination into a utility, the same way a search box in the taskbar changed how people looked things up. Google already owns the browser. A system-wide shortcut extends that reach to every other application on the desktop, and it gives Google a daily touchpoint that does not depend on Chrome.
Skills Replace Gems on November 17 #
Skills let users save instructions they reuse. Jetstream says the feature began rolling out on October 13, 2026. You type a slash followed by the skill name to invoke one, or let Gemini apply a skill in the background when it fits. You can also combine several skills in one request, and a dedicated Gemini Skills page handles management.
Skills replace Gems, Google’s earlier system for saving custom assistants. Automatic migration of existing Gems starts on November 17, 2026, according to jetstream, and Creators Toolbox confirms existing Gems move over. Skills are for users 18 and older. Japan gets the feature at launch.
The practical advice is short. If you rely on Gems, nothing breaks on migration day, but you should open your list before November 17 and note which ones matter. Rename anything ambiguous, since a slash command rewards short and distinct names. The stacking ability is the real upgrade. A “tone” skill plus a “format” skill plus a “source-checking” skill, run together, beats one sprawling prompt that nobody wants to maintain.
Maps, Notebook and the Personal Intelligence Push #
Two features deepen Personal Intelligence, the layer that lets Gemini answer from your own data. The first adds Google Maps. Gemini can now answer questions using your place search history and saved locations, alongside Calendar and Gemini Live. Jetstream gives an example of asking where you went on a trip and getting real-time operating status for those places. Google Japan promoted the feature on Instagram on October 21, 2026, and the rollout is gradual.
The second feature links Gemini to Gemini Notebook. Typing “@Gemini Notebook” in chat lets you add content to a notebook or ask it questions without leaving the conversation. Gemini can also invoke the integration on its own when a prompt looks relevant. It works for personal Google accounts and Google Workspace. The setting is on by default worldwide and off by default only in the EEA.
The EEA carve-out is the tell. Google defaults the integration off where privacy rules are strictest, and that tells us the company sees these features as data-hungry enough to draw regulatory attention. A feature that reads your location history and your notes can feel helpful or invasive depending on the day.
Connectors Pass 40 Apps #
The app connectors list gained Adobe, Strava and Peloton, which pushes the total past 40 connectable apps, according to jetstream.blog. Google renamed the feature to “Connectors” in mid-October 2026 and revamped the interface with a search function. You connect apps through settings, mention them in chat or just ask for what you need.
Availability is the weak point. Connectors work mainly in US English, with some support in other English-speaking regions, Brazil and India. Japan does not get them. A feature that reaches only a slice of the market makes the headline count of 40-plus apps look bigger than the audience that can use it.
The additions are telling. Strava and Peloton are fitness services, and Adobe is creative software. Together they show Google moving beyond email and calendars into hobbies and creative work. For an assistant to feel personal, it needs to know what you do on a Tuesday evening, and fitness data answers that better than an inbox does. The same logic drove our reporting on Minicor’s $3.8M seed round for AI agents that drive legacy desktop apps, where the pitch is that agents win by reaching software that has no clean API.
File Creation and Editing Inside Chat #
Gemini can now create, preview and edit files without leaving the conversation. Jetstream lists documents, PDFs, Google Docs, Sheets, Slides, Markdown and text files, and code. The feature went live on October 29, 2026. Users can open a file full-screen, highlight a passage and request a change, or edit by hand. Gemini can also turn an image into a PDF and refine it on request.
One summary adds a data example: select a table and ask Gemini to convert it into a graph. That is a small trick with a large effect on everyday work, since charting a table is a chore most people avoid.
The platform split is uneven. PDF editing works on web, Android and iOS. Every other format works on the web only for now. Japan gets the feature. If you plan to use it on a phone, expect PDFs and little else until Google widens support.
This feature competes with the document editors people already use. Once Gemini can write, edit and chart inside a single thread, the chat window starts to replace the app switcher. Whether people trust it with a final contract or a board deck is another matter, and the sources report no accuracy data on edits.
Immersive Explorations and Lyria 3.5 #
“Step inside with immersive explorations” targets students. Jetstream says it began rolling out on October 28, 2026, inside the students hub of the Gemini app. It lets learners explore a topic layer by layer, from microscopic systems to deep space. Example topics include Silk Road trade routes and the inside of a jet engine. It supports English, Indonesian, Hindi, Brazilian Portuguese, French and Simplified Chinese. Google plans to add audio narration and direct chat access later. Japan does not officially get it, though people there can use it in English.
Lyria 3.5 is Google’s updated music model. Jetstream dates its announcement to September 4, 2026. Google calls it its highest-quality music generator so far, with more expressive vocals and richer arrangements. Users can choose a genre, pick vocals or instrumental, request short or long tracks and start from purpose-based templates. It runs in the Gemini web, Android and iOS apps, plus the Gemini API, Google Vids, Google Flow Music and Google AI Studio, and it reaches Japan.
Both features show Google using Gemini as a hub for media creation, not only text. The multi-surface release of Lyria 3.5 matters for developers, because API access means third-party apps can generate music with the same model that powers the consumer app.
Shared Avatars and the Privacy Questions #
Shared avatars let you give friends, family and collaborators access to an avatar you made in Gemini, then tag them in prompts. Per jetstream, the result can be images or video featuring both of you, with your likenesses and voices. You can add or remove access whenever you like. Creators Toolbox says sharing requires a Gmail address, recipients can decline, and new images and videos stop once you end sharing while existing work stays.
The limits are tight. The feature covers personal Google accounts for people 18 and older, in English, and excludes the EEA, Switzerland and the United Kingdom. Video generation requires a Google AI plan. The feature first surfaced in a teardown of Google app version 17.54.12 in early September 2026 under the codename Wasabi, and it appeared in a provisional state on iOS in Japanese.
Where the risk sits
Voice and likeness cloning between consenting friends is useful for family videos and risky everywhere else. The safeguards named in the reports are consent through a Gmail invitation, the right to decline and revocable access. The reports do not describe watermarking or abuse reporting for avatar content, so we cannot say how Google handles misuse. That is the question to put to Google before the feature reaches Europe.
Where Each Feature Is Available #
The second table sorts the features by the Japan test that jetstream.blog applied. Seven of 11 reach Japan, and four do not. Japan works as a proxy for non-US markets, though the exclusions for avatars and the Mac app also cover the EEA, Switzerland and the UK.
| Feature | Reaches Japan | Language limit | Plan or account limit |
|---|---|---|---|
| Gemini 4 Argon | Yes | None stated | AI Pro or Ultra |
| macOS app with Antigravity | No | English only | Age 18+, supported regions |
| Windows app | Yes | Same as web Gemini | Windows 10 or later |
| Skills | Yes | None stated | Age 18+ |
| Maps in Personal Intelligence | Yes | None stated | Gradual rollout |
| Gemini Notebook integration | Yes | None stated | Personal and Workspace accounts |
| Expanded connectors | No | Mainly US English | Some English regions, Brazil, India |
| File creation and editing | Yes | None stated | PDF on mobile, rest web only |
| Immersive explorations | No | 6 languages | Students hub |
| Lyria 3.5 | Yes | None stated | Several surfaces |
| Shared avatars | Not officially | English | Personal accounts, age 18+ |
The pattern is clear. Features that act as plain tools, such as Skills, files and music, travel well. Features built on personal data, regional services or likeness cloning stay in English-language markets first. Expect the same sorting every month: the closer a feature sits to personal data, the narrower the first wave.
How the Drop Compares With Rival Assistants #
Google’s package lands in a market where every major lab ships agents, desktop apps and personal-data features. We have no head-to-head benchmark data for Argon, so a numeric comparison would be invention. What we can compare is product shape, using what our site has already reported.
- Agent with approval: Google’s Antigravity asks for approval of a plan, while Anthropic’s Claude Duet uses four approval gates across two plans, as covered in our earlier piece.
- Coding strength: Mistral’s Le Chonk matched a leading rival on a coding-agent index with a score of 48, per our coverage. Google offers no equivalent public score for Argon yet.
- Desktop presence: Google now ships apps for both macOS and Windows with a global shortcut on Windows.
- Personal data: Maps, Calendar, notebooks and 40-plus connectors give Google the widest consumer data surface of the group, since it already runs the underlying services.
Google’s edge is distribution and data it already holds. Its weakness is the absence of published numbers. Rivals that publish scores, even flawed ones, give buyers something to argue about. Silence leaves Google’s claims untested. For a deeper look at the open-model side, see our Le Chonk coding-index report.
Historical Context: From Gems to Skills #
Gemini’s feature set has moved in a consistent direction. It started as a chatbot, added saved assistants in the form of Gems, then added tools, connectors and now agents. Skills formalize what Gems began: reusable instructions that behave like small programs. Replacing Gems with Skills, and migrating them automatically from November 17, shows Google willing to retire its own branding when a cleaner abstraction appears.
The monthly Drops format has its own history. Rather than saving features for a developer conference, Google now ships whenever a piece is ready and bundles later. The October edition proves the habit, since items dated September 4, September 10, October 13, October 28 and October 29 all appear under one label. That cadence rewards patience. Readers who follow the individual release notes know about most of these features weeks before a roundup names them.
The Windows app also closes a long-standing gap. For years Google’s assistant lived in a browser tab or a phone while rivals shipped native desktop clients. Shipping on macOS and Windows puts Gemini on the two platforms where knowledge workers spend their day.
Market Impact: What the Drop Means for the AI Race #
The drop arrives while investors question the returns on AI spending. Our reporting on Big Tech growth cooling, with Google search growth falling from 19% to 12% shows why Google needs Gemini to earn its keep. A search business that grows more slowly has to find revenue elsewhere, and subscriptions to Google AI Pro and Ultra are the obvious candidate.
Look at how the features map to subscriptions. Argon sits behind AI Pro and Ultra. Avatar video requires a Google AI plan. Lyria, files and Skills reach a wider group. The pattern is a free tier that shows the product and a paid tier that unlocks the most capable model and the media features. That structure is a standard freemium ladder, and the October drop adds rungs to it.
For enterprise buyers, the Notebook and Workspace support matter more than avatars. Gemini working inside Workspace accounts, with a default-off setting in the EEA, shows Google designing for compliance teams. The Mac and Windows apps give IT departments something to deploy. None of this proves adoption, because the reports contain no user counts, and we will not guess at them. The competitive pressure also runs in the other direction. When a consumer chat app can write files, run a coding agent and generate music, standalone single-purpose tools lose their easy pitch. Smaller vendors will need to point to depth, price or privacy to justify a separate subscription.
Four Predictions for the Next Drops #
These are our analytical judgments, not reported facts. We base them on the release pattern in the October package.
- Antigravity reaches more platforms. The Mac app launched first, and Windows already has a native client, so extending the agent there looks like the cheapest next step. We expect the restriction to supported regions to loosen slowly.
- Skills absorb connectors. A skill that calls Strava or Adobe is a natural pairing, so expect Google to blend the two systems in a future drop.
- Argon gets published numbers. Pressure from rivals that publish scores will push Google toward benchmark disclosure within a few months, or independent testers will fill the gap for it.
- Europe stays a step behind. The EEA, Switzerland and the UK are excluded from avatars and default-off for Notebook, and we expect that sequencing to continue for personal-data features.
The safest bet is that the next Drops will again mix new launches with items already shipped. Watch the dates in each roundup to separate the two.
What Readers Should Do Now #
Before November 17, review your saved Gems and decide which to keep as Skills. If you code on a Mac and live in a supported English-language region, try Antigravity on a throwaway project first and read every plan before approving it. Check whether your Argon access comes through AI Pro or Ultra before you rewrite any workflow around it.
For privacy, review the Maps and Notebook settings, especially if your account is in the EEA where they default off. Anyone sharing an avatar should remember that a Gmail invite is the only gate named in the reports. For broader context on where this category is heading, browse our AI and machine learning coverage.
FAQ: Gemini Drops October 2026 #
What is Gemini Drops October 2026? It is Google’s monthly feature roundup for the Gemini app. The October edition lists 11 features, including Gemini 4 Argon, Antigravity on Mac, a Windows app, Skills and shared avatars.
When did Google announce it? Jetstream.blog cites a post from the @GeminiApp account on October 30, 2026. Some outlets date it to October 29. Individual features shipped earlier, including the Windows app on September 10 and Argon on October 13.
Do I need a paid plan? Gemini 4 Argon requires Google AI Pro or Google AI Ultra, and avatar video needs a Google AI plan. Reports conflict on whether the other features reach free users, so check the feature’s own page.
What happens to my Gems? Skills replace Gems. Automatic migration of existing Gems starts on November 17, 2026, according to jetstream.blog, and Creators Toolbox confirms existing Gems move over.
Is the Antigravity Mac app available everywhere? No. It supports English only, requires users aged 18 or older and works where Antigravity is supported. Creators Toolbox also lists the EEA, Switzerland and the UK as excluded. Japan does not get it at launch.
How many of the 11 features reach Japan? Seven, according to jetstream.blog and AI Matter In. The Mac app, connectors and immersive explorations do not, and avatar sharing has no official Japan launch yet.
Has Google published benchmarks for Gemini 4 Argon? Not in the coverage we reviewed. The sources describe strengths in programming, finance, law and cybersecurity defense but cite no test results.