Moonshot AI Launches Kimi Work, a Local Desktop Agent Reportedly Running on Kimi K2.6 With a 300-Sub-Agent Agent Swarm
Moonshot AI has launched Kimi Work, a new desktop AI agent that operates locally on your computer. This agent can read local files, control your browser, and perform scheduled tasks, targeting knowledge workers who need direct access to their files and live sessions. Unlike cloud-based agents, Kimi Work keeps data on your device and interacts with your existing environment. It is reportedly powered by the Kimi K2.6 model, a large open-weight Mixture-of-Experts model.
Moonshot AI has released Kimi Work, a new desktop AI agent designed to operate locally on your computer. This tool aims to assist knowledge workers by providing direct access to local files, controlling web browsers, and executing scheduled tasks. It offers a distinct alternative to traditional cloud-based AI agents, keeping data on your device and integrating seamlessly with your existing workflow.
Unlike most previous AI agents that operated on remote servers, Kimi Work runs as a downloadable application. Users provide goals in natural language, and the agent performs actions directly on their machine. Independent reports suggest it leverages Moonshot AI's flagship Kimi K2.6 model, an open-weight Mixture-of-Experts model with a 256K-token context window.
Kimi Work also includes pre-integrated financial data for various markets, removing the need for custom API setups. This allows for direct research and conversion of findings into formats like PowerPoint presentations or Excel spreadsheets. The core advantage of Kimi Work lies in its local execution, offering enhanced data control and direct interaction with the user's computing environment.
The agent is driven by natural language commands and features a cron engine for scheduling tasks. This enables users to automate complex workflows by pairing plain-language tasks with standard cron schedules, such as drafting a daily market briefing at a specific time.
Related articles
Build real agentic apps using CUGA: two dozen working examples on a lightweight harness
CUGA, IBM's open-source Agent Harness, simplifies building agentic applications by handling infrastructure, allowing developers to focus on tools and prompts. It offers pre-assembled components for planning, execution, and state management, significantly reducing development time. CUGA has topped agent benchmarks like AppWorld and WebArena.
OpenAI launches new initiative to help find and patch open source bugs
OpenAI has launched "Patch the Planet," a new initiative in partnership with cybersecurity firm Trail of Bits, to enhance the security of open-source projects. This program aims to assist maintainers in identifying and patching bugs, utilizing OpenAI's AI-powered security tools while reducing the burden on project teams.
PP-OCRv6 on Hugging Face: 50-Language OCR from 1.5M to 34.5M Parameters
Baidu has released PP-OCRv6, an advanced optical character recognition (OCR) model supporting 50 languages. Available on Hugging Face, this version significantly improves accuracy and efficiency across various parameter sizes, from 1.5 million to 34.5 million, marking a substantial leap in multilingual OCR technology.
