What Is Tabbit 1.0
Tabbit 1.0 is an AI-native browser from Meituan's GN06 team with a built-in agent that reads pages, fills forms, and runs cross-site tasks from a single command. The GN06 team launched version 1.0 on June 9, 2026 after a 100-day public beta, during which its agent task success rate climbed from 53.1 percent in March to 91.8 percent. It is aimed at knowledge workers who juggle many web apps.
Core Features
- Agent mode: give a plain-language instruction and Tabbit clicks buttons, fills inputs, and delivers results in a separate tab group.
- Multi-model backbone: switch between LongCat, DeepSeek, Zhipu GLM, and Kimi domestically, or Gemini, GPT-5.2, and Claude Sonnet 4.6 internationally.
- Memory: a callable memory stores your preferences and background so replies adapt to your style over time.
- At-sign context: type @ to pull in the current tab, a screenshot, bookmarks, or a local file without copy-paste.
- Skill library: save reusable prompts as MiaoZhao and call them with a slash command.
Use Cases
- HR teams going from resume screening to a finished PPT in one sentence.
- Researchers pulling facts across dozens of open tabs into a single brief.
- Power users who want one browser that also switches foundation models per task.
Pricing
Tabbit 1.0 is free for core features: chat, web reading, and common MiaoZhao, with Windows and macOS apps live and mobile in testing. Meituan plans a paid subscription for high-frequency agent automation and advanced customization to cover compute cost. For similar workflow tools, see our productivity category.
Our Take
Best for heavy web workers inside the China ecosystem who want an agent that acts, not just answers; the trade-off is that sensitive steps like payments still need your confirmation, and the international model lineup trails US-only rivals. Use it as a doer, not a chatbot.
