Kimi K2.5

Native multimodal model with self-directed agent swarms

Kimi K2.5 is Kimi’s most advanced multimodal AI model to date, delivering state-of-the-art (SoTA) performance in coding, visual understanding, and agent-based workflows. It natively processes text, images, and videos, turning prompts into high-quality outputs like functional code and structured documents. With modes including Instant, Thinking, Agent, and Agent Swarm, up to 100 sub-agents can run tasks in parallel, accelerating complex workflows with power and precision.