Google has updated Gemini Spark more often in the past three months than most software products change in a year, and keeping track of what’s actually new versus what was announced back in May can get confusing fast. Since its debut at Google I/O 2026, Spark has picked up a faster underlying model, a native Chrome integration, a dedicated Gmail address you can email directly, a macOS desktop presence, and a growing list of third-party apps it can now operate inside — none of which existed when the agent first launched in beta.
This guide is built as a running feature log: what Gemini Spark could do at launch, what’s been added since, and what the agent’s full current capability set looks like today. If you already know the basics of Spark and just want the delta — the actual list of what changed — you’ll find that laid out chronologically below, sourced directly from Google’s own release notes and official blog posts rather than secondhand summaries.
We’ll also walk through how each new capability actually works in practice, because a changelog entry like “Chrome auto-browse” or “moves to Gemini 3.7 Flash” doesn’t mean much without context on what it lets you do differently. By the end, you’ll have a clear, verified picture of Spark’s current feature set, what’s changed recently, what’s still limited, and how to actually use the newest additions.
Gemini Spark is Google’s persistent, cloud-based AI agent inside the Gemini app, and as of this update it runs on the newly released Gemini 3.7 Flash model, connects natively to Chrome, and reaches Google AI Pro subscribers — not just the pricier Ultra tier it launched with.
Quick Summary: What’s Changed Since Launch
| Update | What It Added | Date |
|---|---|---|
| Spark launches in beta | Core agent: Tasks, Schedules, Workspace integration, 24/7 background execution | May 19, 2026 |
| macOS app integration | Spark available inside the Gemini for macOS desktop app | June 30, 2026 |
| Gemini 3.6 Flash rollout | Faster, upgraded base model across the Gemini app | July 21, 2026 |
| Chrome auto-browse | Spark can browse the web using your logged-in Chrome sessions | July 30, 2026 |
| Google AI Pro access | Spark expanded beyond Ultra to the $19.99/month Pro tier (US first) | July 30, 2026 |
| 160+ new countries | Broader international availability for Pro subscribers | July 30, 2026 |
| New connected apps | Granola, Otter.ai, Wix, OpenTable, Ticketmaster, Zocdoc, and more | August 12, 2026 |
| Gemini 3.7 Flash upgrade | Spark’s underlying model upgraded again, with better tool use | August 13, 2026 |
Why Trust This Review
Every update listed in this article is cross-checked against Google’s official Gemini Spark release notes, Google’s official blog announcements, and Google’s Gemini model card documentation. We evaluate updates against consistent criteria: actual functional impact, rollout scope (who gets it and where), reliability signals Google has disclosed, and how each change affects real workflows — not just whether Google issued a press release about it. Nothing below is speculative; where a capability is still rolling out or region-limited, that’s stated explicitly.
What Is Gemini Spark? A Quick Refresher
For anyone new to it, Gemini Spark is Google’s answer to the agentic AI category — a tool that doesn’t just chat with you but works on a goal continuously, running on a dedicated Google Cloud virtual machine so it keeps going even after you close your laptop. It launched as an experimental beta at Google I/O 2026, built around three building blocks: Tasks (the goal you give it), Schedules (automated triggers), and native connections into Gmail, Calendar, Docs, Sheets, and Drive.
Since that beta launch, virtually every major update to Spark has either made it faster and smarter (new underlying models), more reachable on the web (Chrome integration), more affordable (Google AI Pro access), or more connected (new third-party apps). The sections below break down each of those in order.
The Complete Gemini Spark Update Timeline
May 2026: Spark Launches in Beta
At Google I/O, Spark debuted as an experimental beta limited to Google AI Ultra subscribers aged 18 and over in the United States. The original feature set included the ability to run Tasks in the background, set up Schedules for recurring automation, and connect to core Google Workspace apps. Google also confirmed at launch that Spark could be triggered via a dedicated Gmail address — you could email a task directly to Spark rather than opening the app — and that progress could be tracked on Android through a system called Halo.
Spark launched built on Gemini 3.5 Flash and Google’s Antigravity agent harness, and Google noted it would eventually support connections over MCP (Model Context Protocol), the same open standard several competing AI agent platforms use for third-party integrations.
June 2026: Gemini Spark Comes to macOS
Spark’s next major milestone was landing inside the Gemini app for macOS, giving Mac users a native desktop presence for the agent rather than only accessing it through a browser tab or phone. At this stage, Spark on macOS could organize folders, pull from local files to help build documents, and run Workspace-based workflows — still limited to English and Google AI Ultra subscribers. Google also flagged that commanding Spark to access local Mac files remotely, from the web or mobile app, was coming soon rather than available immediately.
July 2026: A Faster Model, Then Chrome Integration
Gemini 3.6 Flash rolled out across the Gemini app in mid-July, an upgraded version of the Flash model family aimed at improving performance on document analysis and multi-step projects — a quiet but meaningful improvement for Spark’s underlying reasoning quality.
Then, on July 30, Google shipped what’s arguably the most consequential Spark update to date: direct Chrome browser integration. With permission, Spark gained the ability to use a person’s logged-in Chrome accounts and saved passwords to complete web-based errands that don’t have a native API connection — researching flight options, starting a booking, or scheduling an apartment viewing from saved listings. Google paired this with stated protections against prompt injection attacks and confirmed that sensitive actions like payments are handed back to the user rather than completed autonomously.
The same July 30 update also expanded Spark to Google AI Pro subscribers in the US for the first time — a significant shift, since Spark had been Ultra-exclusive since launch — and extended country availability to over 160 additional markets for Pro subscribers.
August 2026: New Connected Apps and a Second Model Upgrade
On August 12, Google announced a fresh wave of third-party connected apps rolling out to Gemini, several of which extend directly into what Spark can act on. The new categories include:
- Productivity & creativity: Granola, Otter.ai, Wix
- Local & entertainment: Fever, GetYourGuide, Localiza, OpenTable (UK), Ticketmaster
- Music: iHeartRadio, Pandora
- Home, health & lifestyle: Angi, Thumbtack, Zocdoc
Just one day later, on August 13, Google released Gemini 3.7 Flash, and moved Spark onto the new model the same day. According to Google, this update makes Spark more efficient for knowledge work, with improved tool use across Google Workspace apps — specifically calling out gains in consolidating files, drafting emails, and updating status documents across complex, multi-skill workflows.
Independent early benchmarking has also shown measurable gains over the previous Flash generation on coding and tool-use tasks, though Spark’s real-world task completion rates from this specific upgrade haven’t been independently published.
Feature Deep Dive: What Each Update Actually Does
Gemini 3.7 Flash: The New Engine Behind Spark
Spark’s biggest under-the-hood change is its move to Gemini 3.7 Flash, Google’s newest workhorse model, released just three weeks after its predecessor, Gemini 3.6 Flash. Google describes the upgrade as improving Spark’s tool use — meaning how reliably it calls on Gmail, Docs, Sheets, and other connected apps mid-task — along with better output quality on multi-skill workflows that chain several steps together.
In practice, this should show up as fewer dropped steps and more accurate results on tasks like consolidating scattered files into one document or keeping a status tracker updated automatically. The model swap happens automatically in the background; there’s no separate setting to toggle.
Chrome Auto-Browse: Spark Leaves the Google Ecosystem
Before this update, Spark’s usefulness was largely capped by which apps had a native API connection. Chrome auto-browse changes that by letting Spark navigate the open web using your existing logged-in sessions, which means it can now do things like compare options across multiple travel or rental sites, fill out forms, and begin — but not complete — purchases, since payment confirmation is deliberately left to you. This is currently a US-first rollout, and Google has said it plans regional expansion but hasn’t given a firm timeline.
Google AI Pro Access: A Much Bigger Audience
Perhaps the most impactful non-technical update is pricing access. Spark spent its first two months as a Google AI Ultra exclusive — a $99.99 to $199.99 per month commitment. Rolling it out to Google AI Pro subscribers at $19.99 per month opens the agent to a dramatically larger pool of users, and combined with the 160-country expansion, this is the update most responsible for Spark going from “niche premium feature” to something a much wider audience can realistically try.
New Connected Apps: Beyond Google’s Own Products
Until August, Spark’s world was essentially Google’s own ecosystem — Gmail, Calendar, Docs, Sheets, Slides, Drive, Maps, YouTube. The new third-party connections change that meaningfully: Otter.ai and Granola let Spark work with meeting notes and transcripts, OpenTable and Ticketmaster let it act on bookings and reservations, and Zocdoc and Angi/Thumbtack extend it into healthcare and home-services scheduling. This is the clearest sign yet that Google intends Spark to eventually orchestrate tasks across a person’s entire digital footprint, not just their Google account.
How to Access the Latest Gemini Spark Features
- Update your Gemini app. Make sure you’re on the latest version on web, Android, iOS, or macOS so newly rolled-out features actually appear.
- Confirm your subscription tier. Google AI Pro ($19.99/month, US-first for Spark) or Google AI Ultra ($99.99–$199.99/month) is required — Spark isn’t on the free or Plus tiers.
- Check your account type and age. A personal Google Account and being 18 or older are required; work and school accounts aren’t currently supported.
- Turn on Chrome permissions for auto-browse. In Chrome, go to Settings > AI Innovations > Gemini in Chrome to manage whether Spark can use your logged-in sessions for web tasks.
- Connect new apps individually. Head to your Gemini settings to opt in to any of the newly added third-party services — none are connected automatically.
- Start a Task and let the model upgrade work in the background. You don’t need to select Gemini 3.7 Flash manually inside Spark; the upgrade applies automatically to Spark tasks.
Use Cases That Benefit Most From the Latest Updates
- Multi-app research and booking: Chrome auto-browse plus new connected apps like OpenTable and GetYourGuide let Spark handle end-to-end trip and event research that used to require several separate tools.
- Meeting follow-up: With Otter.ai and Granola connected, Spark can pull from meeting transcripts and notes when drafting follow-up emails or status updates.
- Complex document consolidation: The Gemini 3.7 Flash upgrade specifically targets multi-skill workflows like pulling scattered files into one clean document, which benefits anyone managing reports or client files.
- Home and health admin: New connections to Zocdoc, Angi, and Thumbtack mean Spark can now be asked to help find a doctor’s appointment slot or a local repair professional, not just manage inbox and calendar tasks.
Pricing After the Latest Update
| Plan | Monthly Price | Spark Access | What Changed |
|---|---|---|---|
| Google AI Free | $0 | No | No change |
| Google AI Plus | ~$4.99–$7.99 | No | No change |
| Google AI Pro | $19.99 | Yes — new as of July 30, 2026 | Newly added; US-first |
| Google AI Ultra | $99.99 / $199.99 | Yes — original tier | No pricing change; same tier since launch |
Confirm current pricing on Google’s official subscriptions page, since consumer AI pricing has shifted more than once in 2026.
What Hasn’t Changed (Current Limitations)
- Spark is still capped at 15 concurrently running tasks, regardless of subscription tier.
- It remains restricted to personal Google Accounts — work and school accounts still aren’t supported.
- Chrome auto-browse and the newest connected apps remain US-first, with other regions still waiting.
- Google continues to label Spark an experimental, early-development feature, meaning further changes should be expected.
- Spark still shares Gemini’s broader compute-based usage limits rather than having a fully separate quota.
Gemini Spark vs. Its Earlier Version: Before and After
| Capability | At Launch (May 2026) | Current (August 2026) |
|---|---|---|
| Underlying model | Gemini 3.5 Flash | Gemini 3.7 Flash |
| Subscription access | Google AI Ultra only | Google AI Ultra + Google AI Pro |
| Country availability | Select countries, US-focused | 160+ additional countries added |
| Web browsing | Native Google apps only | Native apps + Chrome auto-browse |
| Third-party apps | Google ecosystem only | Otter.ai, Granola, Wix, OpenTable, Ticketmaster, Zocdoc, and more |
| Desktop presence | Web and mobile only | Web, mobile, and native macOS app |
Pros & Cons of the Latest Changes
| Pros | Cons |
|---|---|
| Meaningfully faster, more accurate model with 3.7 Flash | Model upgrades aren’t user-selectable or transparent in-app |
| Chrome integration extends Spark far beyond Google’s own apps | Chrome auto-browse is still US-only |
| Google AI Pro access makes Spark far more affordable | Newest connected apps are limited in number and region |
| New third-party apps genuinely widen practical use cases | Still labeled experimental, so expect further changes |
| 160-country expansion broadens real-world availability | 15-task concurrency limit hasn’t increased |
Alternatives Worth Knowing
If you want to compare Spark’s newest capabilities against other AI agents built for browser automation, scheduling, or cross-platform delegation, it’s worth reviewing a few of TechyUpdate’s other agent breakdowns before deciding where to invest your subscription budget — particularly if Spark’s current US-first rollout for Chrome and app integrations doesn’t fit your region yet.
Read more: Gemini Spark Latest Update
FAQ
What’s the newest update to Gemini Spark?
As of this writing, the newest confirmed update is Spark’s move to the Gemini 3.7 Flash model on August 13, 2026, which Google says improves tool use and output quality on complex, multi-skill workflows.
Is Gemini Spark still Ultra-only?
No. As of July 30, 2026, Spark expanded to Google AI Pro subscribers in the US, in addition to its original Google AI Ultra availability.
What model does Gemini Spark use now?
Spark currently runs on Gemini 3.7 Flash, an upgrade from the Gemini 3.5 Flash model it launched with in May 2026.
When did Gemini Spark get Chrome integration?
Google announced Chrome auto-browse for Spark on July 30, 2026, allowing it to use logged-in Chrome sessions to complete web errands.
What new apps can Gemini Spark connect to now?
The most recent additions, announced August 12, 2026, include Granola, Otter.ai, Wix, Fever, GetYourGuide, Localiza, OpenTable (UK), Ticketmaster, iHeartRadio, Pandora, Angi, Thumbtack, and Zocdoc.
Does the Gemini 3.7 Flash upgrade cost extra?
No. The model upgrade is included automatically for existing Google AI Pro and Ultra subscribers using Spark; there’s no separate charge.
Is Chrome auto-browse available outside the US?
Not yet. It’s currently rolling out in the US first, with Google stating plans to expand to more regions over time.
Can I still use Gemini Spark on a Mac?
Yes. Spark has been available in the Gemini for macOS app since June 30, 2026, and Google has said local file access from mobile and web is planned.
How many tasks can Gemini Spark run at the same time?
Up to 15 tasks can run concurrently, a limit that hasn’t changed since launch.
Did Gemini Spark’s task limit increase with the latest updates?
No, the 15-concurrent-task cap has remained the same through every update so far.
Is Gemini Spark’s email trigger still available?
Yes. Since launch, Spark has supported being triggered by emailing a dedicated Gmail address directly, in addition to using the in-app Tasks interface.
What is Android Halo in relation to Gemini Spark?
Halo is the Android system Google introduced alongside Spark’s launch to let users track an agent’s task progress from their phone.
Does Gemini Spark support MCP (Model Context Protocol)?
Google indicated at launch that Spark would support broader integrations over MCP, with more connections expected to roll out over time.
Will Gemini Spark come to Google Workspace business or education accounts?
Not currently. As of the latest updates, Spark remains limited to personal Google Accounts only.
What’s the difference between Gemini 3.6 Flash and Gemini 3.7 Flash in Spark?
Gemini 3.6 Flash was a general Gemini app model upgrade in July 2026, while Gemini 3.7 Flash, released three weeks later, is specifically credited by Google with improving Spark’s tool use and multi-step task accuracy.
Are there more Gemini Spark updates expected soon?
Google hasn’t announced a specific roadmap, but given the pace of updates since May 2026, further expansion of regions, connected apps, and model upgrades is likely.
Conclusion
In roughly three months, Gemini Spark has gone from a US-only, Ultra-exclusive beta to an agent that runs on a noticeably stronger model, understands the open web through Chrome, reaches a much larger subscriber base through Google AI Pro, and can act across a real mix of third-party services beyond Google’s own apps. Each individual update — the 3.6 and then 3.7 Flash upgrades, Chrome auto-browse, the Pro-tier rollout, the new connected apps — might look incremental on its own, but stacked together they represent a genuinely faster pace of improvement than most competing AI agents have shown over the same stretch.
What hasn’t changed is just as informative: the 15-task concurrency limit, the personal-account-only restriction, and the experimental label are all still in place, and several of the newest capabilities remain US-first. That’s a reasonable trade-off for a product moving this quickly, but it does mean international users and power users chasing higher concurrency will still hit real walls.
Our take: if you already use Gemini Spark, the recent updates are worth actively testing rather than ignoring — particularly Chrome auto-browse and the new connected apps, which meaningfully expand what the agent can be trusted with. If you haven’t tried Spark yet and you’re on Google AI Pro, this is the best the agent has ever been, and the most accessible it’s likely to remain before the next round of changes arrives.
1 Comment
Pingback: How to Use the New Gemini Spark Features After the Latest Update