Artificial intelligence is extending its application boundaries into the browser domain. On June 9th, the GN06 team under Meituan officially released the AI-native browser Tabbit 1.0 version. This version not only launched simultaneously on Windows and macOS platforms, but its mobile version has also entered full-scale testing, marking the product's transition from the public beta phase to a mature application stage.
Looking back at Tabbit's development path, its evolution speed has been astonishing. Since its public beta launch on March 2nd, the product has undergone more than 100 function iterations. Its core capability - the task execution success rate of the Agent (intelligent agent) - has risen from an initial 53.1% to 91.8% currently. This data not only demonstrates the technical team's problem-solving capabilities but also reflects users' real demand for AI's deep involvement in complex workflows. Backend data from May showed that the average monthly Token usage per user has exceeded 8.53 million, indicating that Tabbit is becoming a powerful assistant for many users handling frequent and intensive office tasks.
In terms of functional logic, Tabbit 1.0 aims to break the common "model binding" pain points in current AI applications. This version includes several leading domestic large models such as DeepSeek, Zhipu GLM, Kimi, and LongCat, and supports real-time access to new model interfaces. This "learning from the best" approach allows over 60% of active users to flexibly call or compare different model performances according to various business scenarios, thus obtaining better inspiration and decision support.
To further enhance the product's stickiness, Tabbit 1.0 introduced a brand-new "memory function." It can continuously learn and accumulate user preferences, background information, and work habits, building a personalized "callable memory," significantly reducing repetitive instructions and inefficient interactions, making AI responses more accurate and considerate. At the launch event, Liu Jiong, the person in charge of Tabbit, demonstrated the tool's practical capabilities in complex scenarios: whether it was a college student organizing a thesis workflow or an HR professional screening resumes across software and generating presentations, users could complete originally complicated collaborative processes with just one instruction.
In terms of business model, Tabbit adopts a tiered strategy that balances accessibility and professionalism. The standard version remains permanently free, offering sufficient conversation, AI-generated images, and ten-thousand-word summary quotas each week; while for high-frequency users, the professional version increases the task execution limit by more than 10 times, with highly affordable pricing, requiring only 9.9 yuan per week.
As an AI entry point evolved from a web-based tool, Tabbit's future ambition lies in building an open "Skill (skill) ecosystem." The team revealed that they will later launch a "Tip Square" and support programs for creators. In the increasingly competitive AI track, this open strategy that avoids being locked into a single model vendor may be the key for Tabbit to find a differentiated competitive advantage.
