Can Open Models Keep Up With OpenAI and Anthropic? | Nathan Lambert

AI model development hinges on the strategic allocation of compute, data, and talent, with closed-source frontier labs like OpenAI and Anthropic currently leading through massive capital investment. While these labs dominate by scaling proprietary infrastructure and inference capabilities, the open-source ecosystem provides a vital, albeit slower, path for innovation and democratization. Distillation—training smaller models on the outputs of frontier systems—remains a critical, if technically debated, lever for fast-following competitors, including Chinese labs. The "frontier" itself is less a hidden secret than a byproduct of extreme resource scaling. Initiatives like Trillium aim to demystify these development processes by providing open-source research environments, ultimately helping society better prepare for AI-driven risks and ensuring that the benefits of advanced intelligence are not confined to a few dominant players.