Perplexity Launches Frontier Research With 10,000+ GPUs
TL;DR
- Perplexity opened research.perplexity.ai with the stated goal of building an AI that can conduct research on its own, backed by more than 10,000 GPUs.
- The agenda spans six pillars: multi-agent orchestration, pre-training, continual learning, reinforcement learning, systems, and evaluations.
- The site lists 19 arXiv papers as sample prior work and promises open releases of papers, data, models, and code on GitHub and Hugging Face.
Perplexity on Monday launched a new research arm aimed at building what it calls "an automatic research scientist, an AI system that can conduct research on its own," backed by more than 10,000 GPUs and a roster of 19 arXiv papers offered as the group's starting evidence base.
The new site at research.perplexity.ai organizes the work under six pillars: multi-agent orchestration, pre-training, continual learning, reinforcement learning, systems, and evaluations. Andrew Gordon Wilson is named as "Research Lead, Perplexity Frontier Research," with hiring open in San Francisco, Palo Alto, New York and London and a December 5 priority deadline, per reporting from Cryptobriefing.
Perplexity says it wants evaluations that measure "research ability and extrapolation, rather than only recall," and commits to releasing papers, blog posts, data, models and code openly on GitHub and Hugging Face. Of the 19 arXiv papers the site offers as sample work, 11 are dated 2026, five 2025, two 2024 and one 2022.
The announcement names no commercial partners, no first-system release date, and no external benchmark the automated scientist will be measured against.
Shared on Bluesky by 1 AI expert
-
We’re excited to announce Perplexity Frontier Research! research.perplexity.ai We will be openly advancing the frontier of models, data, and algorithms. Join us in re-imagining multi-agent collaboration, architecture des…
View on Bluesky →
Originally reported by research.perplexity.ai
Read the original article →Original headline: Frontier Research | Perplexity