Reporting by |
Editing by | NUPIAO
Meituan’s AI-powered browser, Tabbit, is finally putting a price tag on heavy usage.
On June 9, Meituan’s GN06 team dropped the official V1.0 release of Tabbit—right on the 100-day mark since its March 2 public beta. Alongside the launch came the monetization plan: the Standard tier stays free forever, while Pro comes in at ¥9.9 per week, targeting power users who live in conversations, image generation, and agent workflows.
And that’s what makes this moment so pivotal. Over the past year, the AI app landscape has shifted hard from growth-at-all-costs and free trials into real-world commercial testing. Unlike traditional web apps, slapping more users on an AI platform doesn’t magically lower your bills. Every chat prompt, generated image, or agent task burns through tokens and eats up compute. Figuring out how to keep a free entry point open while actually covering those costs? That’s the billion-dollar puzzle every consumer-facing AI product is trying to solve right now.
Tabbit’s playbook is pretty straightforward: keep the core experience totally free, then charge the absolute power users who drain the most resources. You’ll still get plenty of room for chats, image gen, and agent tasks on the Standard plan. Hit the cap? No worries—you can keep going using a designated, budget-friendly fallback model. Upgrade to Pro, and you unlock heavier quotas plus advanced features.
At ¥9.9 a week, the Pro tier actually undercuts a lot of the heavy hitters in the AI space right now. Liu Jiong made it clear during the press call that this isn’t some fleeting launch promo or subsidy trap. Any future pricing shifts will be tied directly to real-world factors like inference costs and actual user demand.
According to a Tabbit product manager speaking to NUPIAO, this split comes down to raw consumption patterns. During the beta, the team noticed a massive gap between casual scrollers and power users. A tiny slice of super-users was burning through way more tokens than their numbers would suggest, effectively subsidizing everyone else’s bill.
It’s also perfectly in line with where the rest of the industry is heading. Back in May, Doubao quietly rolled out paid tiers on the App Store—offering monthly plans at 68 RMB, 200 RMB, and 500 RMB. Doubao’s official response? Free access will always be here; the paid stuff is just about layering in premium perks on top.
Here’s where Tabbit really sets itself apart: it’s not just another chat interface bolted onto a sidebar. It’s a full-blown AI browser. You’ve got direct access to China’s top-tier models, plus built-in agent execution. Want to jump between DeepSeek, Kimi, Zhipu GLM, Doubao, Tongyi Qianwen, MiniMax, and LongCat without leaving the page? Done. Plus, the AI can actually hop across active tabs, local files, and cloud docs to get work done.
Liu Jiong admitted they weren’t even sure if juggling multiple models would catch on before launch. Beta data quickly proved them wrong: over 60% of users actively switch models mid-session, averaging around 2.17 different engines per user. Whether you’re coding, hunting down answers, or crunching translations, that flexibility is exactly what separates Tabbit from the walled gardens pushed by individual AI companies.
Beyond that, Tabbit is betting big on a bigger idea: browsers aren’t dead. They’re still one of the most powerful frontiers for AI integration.
In a March interview with NUPIAO, Liu opened up about how he started mulling over AI browsers back in late 2022. When chatbots were all the rage, he saw a different opportunity. Most people spend way more time consuming and navigating content than actually writing it. And let’s be real—the browser market dwarfs both note-taking apps and online docs combined.
As NUPIAO previously reported, Tabbit went from zero to hero fast. Official kickoff in August 2025, dev sprint started in September, first prototype shipped by November, internal QA wrapped up in December, and it hit the public beta right after Chinese New Year.
What actually gave the green light? GUI agents getting seriously good. Back in 2024, automating clicks and form fills was still pretty buggy. But by early 2025, agents could reliably read interfaces, tap buttons, type into fields, and navigate pages without falling apart. Suddenly, the browser wasn’t just a viewing window—it became a legitimate command center.
Tabbit’s roadmap actually breaks down browser evolution into four clear waves: Gen 1 was just typing URLs. Gen 2 added search bars. Gen 3 slapped AI chat widgets on top. Gen 4? That’s where Tabbit lives—combining navigation, search, conversation, and actual execution into one seamless workflow.
This latest push doubles down on action. The beta agent handled standard tab interactions—clicking elements, auto-filling forms, scraping data. Now in V1.0, it’s reading local files, spitting out multi-format documents, and playing nice with popular third-party agent plugins.
But here’s the real hook: Tabbit wants to transform the browser from a passive info portal into an active AI workstation. Think about it—real work rarely happens in just one place. You’re juggling live web tabs, spreadsheets sitting on your desktop, shared docs, and whatever file format your boss just emailed you. Tabbit leverages the browser’s native grip on tabs, cross-site context, and local files to drop the AI right into the middle of your actual workflow.
Whether that thesis holds water remains to be seen. Then again, the major chatbot players are already migrating outward, pushing into dedicated desktop apps, browser extensions, and agent hubs left and right.
According to reports from The Information, OpenAI is currently stitching ChatGPT, Codex, and its browsing tool Atlas into a single unified desktop client. Google already baked Gemini straight into Chrome, letting it pull context from your open tabs to answer questions, summarize threads, and run comparisons. Even Apple doubled down on Apple Intelligence at WWDC, tightly coupling it with native apps and Safari. The race to own the workspace is definitely on.
Interestingly, when asked who his real competition is, Liu pointed straight at Chrome rather than the newer AI-native browsers. Why? Because startups like Tabbit are still relatively small fish. The massive, entrenched user base is still cruising on legacy browsers. Most folks aren’t looking to overhaul their entire browsing habits—they just want AI smarts layered on top of what they already know.
That mindset explains why Tabbit refuses to pigeonhole itself into a niche demographic. A product manager told NUPIAO that the beta crowd has been surprisingly diverse: hardcore AI tinkerers, students cramming for exams, corporate professionals, and total newcomers who’d never touched an AI app before.
That positioning carves out a unique lane. Tabbit isn’t trying to dethrone pure-play chatbots, nor is it fighting Safari or Edge head-on. Instead, it’s doubling down on three things: agnostic model switching, deep contextual awareness, and actionable automation. Pair that with the vertical sidebar, quick-action skills, and a suite of custom tweaks, and you’ve got a solid blueprint for differentiation.
Then there’s the question of guardrails. Liu emphasized that safety and control are non-negotiable. For anything touching logins, form submissions, or payments, Tabbit forces a manual confirmation step. Most of the heavy lifting for these updates runs securely in the cloud—think API calls and Python script execution—so it won’t accidentally mess with your local machine.
Of course, the elephant in the room is Meituan. Despite the autonomy Liu describes, you can’t ignore the parent company’s shadow.
GN06 actually traces its roots back to Beyond Infinite, the lab founded by Wang Huiwen. After Meituan acquired the team in 2023, the core model engineers merged into Meituan’s main Large Model squad, LongCat. Meanwhile, the application developers reorganized into GN06 to keep pushing forward on true AI-native products.
According to Liu speaking to NUPIAO, GN06 operates as a standalone unit. Their mandate isn’t to bolt AI onto Meituan’s core delivery services—it’s to hunt for greenfield opportunities in native AI apps. With existing internal tools like “Xiao Mei” and “Xiao Tuan” already handling business-specific workflows, GN06 has the freedom to chase a more independent vision.
From an org chart perspective, GN06 runs as its own profit center within Meituan, reporting straight to an S-team executive. Liu notes that decision-making stays remarkably lean. Leadership will throw out suggestions and strategic north stars, but they won’t micromanage feature rollouts or slap unrealistic short-term KPIs on the table.
When the microphones turned to potential Meituan ecosystem tie-ins during the June 9 press briefing, Liu shut that down pretty clearly. There are no current integrations with Meituan’s local life services. Adding food delivery or ride-hailing contexts down the line? He called it a “cherry on top,” stressing that it’ll only happen if the product is truly ready and the goals actually align.
All of this leaves Tabbit in a fascinating limbo—part incubated mega-corp project, part scrappy startup. As an independent AI browser, it’s not going to overnight steal Chrome or Safari’s thunder, nor can it lean heavily on Meituan’s massive traffic funnel. A far more grounded strategy? Carve out a loyal niche among power users doing heavy office work, document crunching, and web automation. Nail those daily workflows, prove the value proposition sticks, and the broader market adoption will follow naturally.
