Moonshot AI

Latest AI news, models and releases from Moonshot AI. ['GLM 5.2', 'GLM-5.2', 'Hermes', 'K2', 'K2.6', 'K2.7', 'K3', 'K3-mini', 'K3-MoE', 'Kimi', 'Kimi 2.6', 'Kimi-2.6', 'Kimi 3', 'Kimi 3T', 'Kimi Hosted Agent', 'Kimi k1.5', 'Kimi K1.5', 'kimi-k2', 'Kimi K2', 'kimi-k2-0711-preview', 'Kimi K2.5', 'Kimi-K2.5-1T-A32B', 'Kimi K2.6', 'Kimi K2.7', 'Kimi K2.7 Code', 'Kimi K2 Thinking', 'Kimi K3', 'Kimi-K3', 'Kimi K3.5', 'Kimi K4', 'Kimi K5', 'Kimi Linear', 'Kivine', 'Krea 2', 'Krea 2 Turbo', 'Mamba', 'Mamba-2', 'Mamba-3', 'MiniCPM-RobotManip', 'MiniCPM-RobotTrack']

Official site ↗

Models 🇷🇺

Moonshot AI releases weights of popular Kimi-K3 — nearly on par with market leaders but much simpler to run

Chinese company Moonshot AI has open-sourced the weights of its Kimi-K3 model. It outperforms previous OpenAI and Anthropic models while costing 2-3 times less to operate, with cache hit rates up to 90% for coding tasks. The model uses mixed precision formats MXFP4 and MXFP8, and is optimized for less powerful hardware like Nvidia H20.

Moonshot AI

OpenAI

Anthropic

3DNews28.07 · 17:02

AI Safety 🇷🇺

Chinese AI models start calling themselves Claude and mimicking American bot behavior

Independent researchers found that leading Chinese AI models like Z.ai GLM 5.2 and Moonshot AI Kimi K3 sometimes identify themselves as Claude, an Anthropic chatbot. In GLM 5.2, adopting Claude's persona weakens Chinese censorship mechanisms, raising safety concerns.

Moonshot AI

Alibaba/Qwen