Moonshot AI unveils Kimi K3

The rapid approach of frontier-class open-source LLMs like Kimi K3 is accelerating the commoditization of intelligence, turning raw tokens into a cheap utility and shifting the real competitive moat to hardware infrastructure and proprietary application-level workflows.

* Commoditization of Tokens: With open-weight models scaling to trillions of parameters, the marginal cost of intelligence is dropping exponentially, signaling a future where tokens are priced and consumed like water or electricity.

* Infrastructure Barriers: While weights are open, running a 2.8-trillion-parameter MoE model at scale requires massive, highly optimized compute infrastructure, keeping smaller players reliant on API providers or specialized hardware clusters.

* Shifting Value Moat: When base model capabilities become a commodity, value shifts entirely up the stack to proprietary data flywheels, domain-specific tooling, and seamless UX.