Moonshot AI Quietly Launches Next-Gen Open Model: Kimi K2.5

‍‌​Moonshot AI hAs officially rolled out its latest open model, Kimi K2.5, via a silent update to its web-based chat interface. The upgrade—deployed server-side—has seamlessly replaced the previous K2 model without requiring any manual intervention from users. All web users now automatically benefit from enhanced response speed, stronger reasoning capabilities, and improved multi-turn dialogue stability.

Although no standalone download or dedicated access point is avAIlable for K2.5, the new model is already live across Moonshot’s entire web platform. According to the company, Kimi K2.5 represents the most capable version of Kimi to date, achieving top scores among all open-source models on key agent benchmarks such as HLE (Humanity's Last Exam), BrowseComp, and DeepSearchQA.

Built on a native multimodal architecture, Kimi K2.5 natively supports both visual and textual inputs, integrating advanced vision understanding, reasoning, coding, and agent capabilities into a single unified model.

Yang ZhiLin, Founder and CEO of Moonshot AI, stated: “We’ve completely rebuilt our reinforcement learning infrastructure and fine-tuned our training algorithms to ensure the model delivers exceptional efficiency and performance.”