We are excited to launch Perplexity Frontier Research, a new initiative to advance the frontier of algorithms, models, and evaluations, in the open.
We aim to build an automatic research scientist, an AI system that can conduct research on its own. Research is a loop that involves asking the right questions, hypothesizing, designing experiments, analyzing results, navigating ambiguity, and updating beliefs. In contrast to black-box predictions in narrowly defined settings, this effort is designed to produce new interpretable knowledge that advances our understanding across a broad range of applications.
This ambition inspires new algorithms, models, and evaluations. We will be particularly considering six pillars: multi-agent orchestration, pre-training, continual learning, reinforcement learning, systems, and evaluations. Within each of these areas, we will explore a variety of foundational questions. For example, can we propose new, more efficient architectures? Can we synthesize data that enables broader generalization? Under what conditions is multi-agent collaboration most effective? Can we create evaluations that measure research ability and extrapolation, rather than only recall?
This is an environment where one has the opportunity to advance the frontier of research, and share results openly. We will release papers, blog posts, data, models, and code. Our research will also be supported by considerable computational resources, in excess of 10,000 GPUs. And our findings shape a product used by millions of curious people, the most challenging test of all.
There are several ways to get involved, including full-time positions and our research fellowship program. Please see our frontier research website for more information about the initiative and research agenda.












