Chinese AI labs, led by Moonshot AI's newly released Kimi K3, are giving away powerful "open-weight" models for free, intensifying competition with closed US systems from OpenAI, Google and Anthropic. Kimi K3 reportedly matches or beats some of the best American models at a fraction of the cost, and Moonshot's decision to release its weights openly while clearly targeting US users has deepened concerns in Silicon Valley about whether proprietary models can keep their dominant position as capable free alternatives spread.

Open-weight models let developers inspect, run locally and customise systems without depending on a single provider, and are typically far cheaper than closed equivalents, though they are not truly open source since training data, code and architecture usually remain private and licences often restrict use. Experts note that giving away weights does not mean giving away a free service, as companies can still profit from computing infrastructure, engineering, hosting and support, or from broader demand for cloud and chip services. Releasing weights can also build an ecosystem of tools around a model, entrenching its use over time, a strategic advantage now pushing Western AI giants to reconsider how much they keep locked away.