Released Kimi K3, a 1.56 trillion parameter MoE model

Open-weight release of Kimi K3 scheduled to offload compute demand

White House accused Moonshot AI of distilling Anthropic's Fable model to train Kimi K3

Paused new subscriptions for Kimi K3 due to GPU capacity limits

Moonshot AI's Kimi K3 achieved 89.2% on SWE-bench Front-End, beating GPT-4o, Claude 3.5 Sonnet, and Gemini 2.0 Pro

Released Kimi K3, a 2.8T-parameter MoE model with 1M token context window

Trained DeepSeek-V3 in ~2 minutes on MLPerf v6.0, beating AWS record by 43%

First cloud provider to validate Nvidia Vera Rubin NVL72

First cloud provider to validate and deploy Nvidia Vera Rubin NVL72 at rack scale

Brought up first Vera Rubin NVL72 system