Simon Willison's piece highlights a critical, under-discussed tension in the current model race: the decoupling of parameter count from both utility and cost-effectiveness. The launch of Kimi K3, with its significant price hike, suggests Moonshot AI is betting that the market will pay a premium for perceived top-tier performance, even from a Chinese lab. In our view, this is a risky strategy that may test the loyalty of developers who have grown accustomed to lower-cost, capable models from the region.
The real story here isn't the pelican benchmark or the parameter crown; it's whether the industry's valuation of model performance is entering a phase where brand perception and benchmark optimization could start to outweigh open-weight accessibility as the primary competitive lever.