Can open-source AI systems outperform closed ones? - Read about KIMI K2 Thinking
Can open-source AI systems outperform closed ones? Thanks to the newest model from Chinese AI research lab Moonshot AI, the answer to that hotly debated question may finally be yes. Kimi K2 Thinking is an open-source agent that uses a mixture-of-experts architecture and outperforms frontier models from OpenAI and xAI on key benchmarks like Humanity’s Last Exam and BrowseComp, according to Moonshot AI. Kimi K2 Thinking is a powerful AI model — here’s a breakdown of what it is, how it works, and why people are excited about it (and some of its caveats). What Is Kimi K2 Thinking? Developer & Origin Created by Moonshot AI . Wikipedia +2 kimik2.com +2 It’s part of the Kimi model series; specifically, the “Thinking” version is made for deep reasoning and agent behavior. Architecture It uses a Mixture-of-Experts (MoE) architecture: total of ~ 1 trillion parameters , but only ~ 32B active parameters per token. Very large context windo...