We are now in the era where we’ll discover whether open models can match the capabilities of closed labs. The clear answer is no, they won’t. This response essentially means they won’t maintain pace or standards across all aspects. This perspective rules out the common forecast that open models will fully catch up, in a scenario where all models reach saturation and open and closed systems grow ever more alike. Navigating this situation, it’s clearly quite uncertain when a longer-term, stable equilibrium of capabilities will emerge. This is a highly intricate dynamic, with our primary focus being the capability gap between models. At the same time, this gap is intertwined with shifting trends in how open models are funded, who is building them, how techniques like distillation enable rapid iteration across new domains, the risk of regulation constraining the open-source AI ecosystem, and, ultimately, who actually adopts open models. The capabilities gap is merely one signal within a complex web of forces that are reshaping both supply and demand. In many cases, clear demand for open models — from countless individuals, organizations, and governments who want or need them — remains largely disconnected from supply. The availability of supply is entirely determined by economic factors.
Interconnects AI