Kimi K2.6

Moonshot AI's latest frontier model, building on the Kimi K2 family's 1T-total / 32B-active MoE architecture with 262K token context. Optimized for long-horizon agentic coding: demonstrated sustained execution across 4,000+ tool calls over 12+ hours with iterative self-improvement. Coordinates 300 sub-agent swarms executing 4,000 coordinated steps simultaneously for parallel document analysis and multi-format output.

SWE-Bench Pro: 58.6%, Terminal-Bench 2.0: 66.7%, LiveCodeBench v6: 89.6%, MathVision w/ Python: 93.2%, DeepSearchQA: 92.5 F1. Supports proactive 24/7 agent operation via OpenClaw and Hermes. Available via Kimi.com, Kimi App, API (platform.kimi.ai), and Kimi Code CLI.