Kimi K3: World's First Open 3T-Class AI Model Launched
Original: Kimi K3: Open Frontier Intelligence
Why This Matters
First open 3T-class model release marks a significant milestone in open-source frontier AI scaling.
Moonshot AI launched Kimi K3, a 2.8-trillion-parameter open model with native vision, 1M-token context, and MoE architecture. It is the world's first open 3T-class model, available on Kimi.com and API. Full weights release planned by July 27, 2026.
Moonshot AI has introduced Kimi K3, its most capable model to date and the world's first open model in the 3-trillion-parameter class, with 2.8 trillion total parameters. The model is built on two novel architectural innovations: Kimi Delta Attention (KDA) and Attention Residuals (AttnRes), designed to improve information flow across long sequences and model depth. It uses a scaled Mixture of Experts (MoE) setup, activating 16 out of 896 experts via a Stable LatentMoE framework, yielding approximately 2.5× improvement in scaling efficiency compared to its predecessor, Kimi K2. Kimi K3 supports native vision capabilities and a 1-million-token context window, making it suited for long-horizon coding, knowledge work, and reasoning tasks. The company notes that while Kimi K3 trails the most powerful proprietary models — Claude Fable 5 and GPT 5.6 Sol — it outperformed all other tested models in their evaluation suite. The model is available immediately on Kimi.com, Kimi Work, Kimi Code, and the Kimi API, with max thinking effort enabled by default. Low- and high-effort modes are planned for future updates. Full model weights are scheduled for public release by July 27, 2026, with a full technical report to follow.