Moonshot AI Releases Kimi K2.6: Open-Source Model Matches Claude Opus 4.6 on Coding Benchmarks

Beijing-based Moonshot AI released Kimi K2.6, a one-trillion-parameter open-weights model that matches GPT-5.4 and Claude Opus 4.6 on multiple benchmarks. The model features 262K context length, native vision through MoonViT, and can orchestrate swarms of up to 300 sub-agents while supporting 12+ hour coding sessions with 4,000+ tool calls. Released under a Modified MIT License, it represents a significant challenge to closed-source frontier models.