> model except to limit its effectiveness in developing frontier LLMs
Does this imply that they're actively using it for their frontier development and that it's very effective?