Nathan Lambert

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

Nathan Lambert, Lisha Li

Leading frontier laboratories like OpenAI and Anthropic leverage massive compute and talent advantages to lead the AI sector, while Chinese competitors rely on model distillation and open-weight strategies to bridge the gap despite compute shortages. To counterbalance this centralization, the nonprofit Trillium aims to demystify recursive self-improvement by releasing verifiable open-source RL environments and transparent documentation of training mechanics. This dual trajectory of proprietary scaling and open research seeks to foster specialized open models for niche industries while ensuring broader societal preparation for advanced AI capabilities.