What happened
Perplexity published “Introducing Perplexity Frontier Research” on 16 November 2026. The post describes a new initiative “to advance the frontier of algorithms, models, and evaluations, in the open”, and gives its mission as autonomous discovery, with the goal of an “automatic research scientist”. The post is dated 16 November 2026 without a time; we read it the same day.
Key details
- Research pillars: the post lists six areas: multi-agent orchestration, pre-training, continual learning, reinforcement learning, systems and evaluations.
- Compute: Perplexity says the initiative has compute “in excess of 10,000 GPUs”.
- Openness: the company says it plans to release papers, blog posts, data, models and code.
- People and programs: the post points to a research site, a fellowship page and open job listings.
- Not stated: pricing, licenses, model sizes, release dates for any model or dataset, and any product change for Perplexity users.
Why it matters
The announcement signals that Perplexity, known for its answer engine, wants to do its own model and evaluation research rather than only build on others’ models. For now it is a statement of mission and resources, not a release: there is no model, benchmark or paper to evaluate yet.
Readers should treat the GPU count and the “automatic research scientist” goal as the company’s own claims. The useful follow-up is whether the promised models, data and code actually appear, and under what licenses.
Disclosure: Claude, made by Anthropic, was one of the AI tools used to research and draft this article. Perplexity is not affiliated with Deyron Labs; all claims are Perplexity’s own.
Sources
- Introducing Perplexity Frontier Research Primary source , Perplexity Research, 16 November 2026