Chinese artificial intelligence startup Moonshot on Friday launched Kimi K3, a 2.8 trillion-parameter open-weight model built on proprietary Kimi Delta Attention and Attention Residuals architecture.
The model, featuring native vision capabilities and a 1-million-token context window, is the world's first open 3T-class model designed for long-horizon coding, knowledge work and reasoning, according to the blog post.
Moonshot said that while Kimi K3 trails behind Claude Fable 5 and GPT 5.6 Sol, it demonstrated frontier-level performance across its evaluation suite to consistently perform better than other tested models.
The full model weights and a technical report will be released by July 27.