With the release of its new large model K2.5, Moonshot AI is reaching a milestone in its global expansion. Recent data shows that the company's overseas revenue has exceeded domestic revenue for the first time, marking a major breakthrough for domestic large models in the international commercial arena.

Global Growth: Paid User Base Quadruples

Within just a few days after the K2.5 model update, Kimi's global performance has been impressive:

Sharp Increase in Users: The number of global paid users has quadrupled in a short period.

High Popularity: On the well-known large model aggregation platform Openrouter, Kimi's popularity has risen to third place, following Claude Sonnet 4.5 and Gemini 3 Flash.

Explosive Revenue Growth: Since November 2025, its overseas API revenue has grown fourfold, with monthly growth rates of paid users both domestically and internationally exceeding 170%.

Core Technology: Native Multimodal and "Agent Swarm"

Kimi K2.5's strong performance stems from a complete innovation in its underlying technology:

Native Architecture: It uses a native multimodal architecture, covering visual understanding, code generation, and complex thinking modes comprehensively.

Performance Benchmarks: It has achieved open-source SOTA levels in multiple benchmark tests such as HLE and BrowseComp, with some metrics even surpassing top closed-source models like GPT-5.2.

Agent Swarm Scheduling: It can simultaneously schedule 100 Agents to handle about 1,500 steps, improving the efficiency of large-scale information collection by 3 to 10 times.

Product Evolution: From Chatbot to "Productivity Engine"

Yang Zhilin, founder of Moonshot AI, pointed out that the team is expanding the model's capabilities through a new dimension called Agent Swarm. In terms of product positioning, Kimi is firmly shifting toward a consumer-end productivity tool:

Brand Unification: The internal testing product OK Computer has been officially renamed Kimi Agent, enhancing general office capabilities.

Deep Editability: Significant R&D efforts have been invested in complex scenarios such as PPT element splitting and editing, generating front-end code from screen recordings, and automatic Excel modeling.

Strategic Focus: The company has clearly defined its business boundaries within the large model layer, logic layer, and Agent layer, focusing on high-value productivity workflows in research, office work, and development.

By maintaining a top-tier global foundation model and deepening Agent capabilities, Kimi is successfully building a globally competitive AI product with unique advantages.