Moonshot AI has open-sourced Kimi K2.6, a new model with state-of-the-art coding, long-horizon execution, and agent swarm capabilities. Key highlights include: demonstrated 12-13 hour autonomous coding runs with thousands of tool calls (e.g., optimizing a Zig-based inference engine to 193 tokens/sec, and achieving a 185% throughput improvement on a financial matching engine); coding-driven UI/full-stack generation from simple prompts; an upgraded Agent Swarm scaling to 300 sub-agents across 4,000 coordinated steps; proactive background agents (OpenClaw) capable of 5-day autonomous operations; and a new 'Claw Groups' research preview enabling heterogeneous human-agent collaboration. Benchmark results show competitive or leading performance vs. GPT-5.4, Claude Opus 4.6, and Gemini 3.1 Pro across agentic, coding, reasoning, and vision tasks.
Table of contents
Long-Horizon Coding Coding-Driven Design Agent Swarms, Elevated Proactive Agents Bring Your Own Agents Benchmark Table Footnotes Sort: