Keep Announces Comprehensive AI Ecosystem Launch for First Half of 2026

On March 25, fitness technology company Keep (03650) released its full-year financial results for the period ending December 31, 2025. During the reporting period, Keep achieved revenue of 1.637 billion yuan and an adjusted net profit of 25.22 million yuan under non-IFRS standards, marking its first annual profit. In 2025, Keep established a strategic shift from being a "content platform" to an "AI-driven fitness and health ecosystem." The company stated that significant progress has been made in its long-term AI strategy, including key iterations in underlying architecture and phased product implementation, laying a solid foundation for strategic transformation. During an earnings call, Keep's founder and CEO, Wang Ning, announced that the company expects to launch the Keep AI Suite in the first half of 2026. This suite will include Keep's self-developed fitness and health large model, Keepace.ai, the fitness and health assistant Keepclaw, and the Keep AI Skill Plaza, completing the comprehensive layout of Keep's AI ecosystem. The initiative aims to provide users with a professional AI-driven fitness and health experience through Skills, Agents, and Models. Specifically, Keep is building core barriers in the fitness and health sector using a "4+1" paradigm: four technical foundations and one business engine. The Model layer leverages billions of real exercise data points accumulated over a decade, with the Keepace.ai vertical large model serving as the fitness core. The Agent layer develops an all-weather proactive AI agent, breaking the passive habit of "users seeking applications." The Skill layer enables a boundaryless skill ecosystem, providing millisecond-level visual error correction and hardware interconnection capabilities. The Benchmark layer establishes industry authority through open-source evaluation standards. Building on this foundation, the industrial-grade AIGC "infinite flow" content engine can break the impossible triangle of high quality, high output, and low cost, achieving zero marginal cost coverage for long-tail demands and real-time dynamic assembly, directly converting computing power into user experience and corporate profits.