Perplexity's Multiagent feature confirmed available to Max and Enterprise Max web users
Perplexity's Frontier Research division launch
Latest: Perplexity Launches Frontier Research Division and Multiagent Feature
Perplexity is building infrastructure and products aimed at autonomous AI research, combining a dedicated compute-backed research division with a multi-model agent coordination feature shipped to users. The SPACE red-teaming findings point to practical limits in how security boundaries for computer agents are currently evaluated.
When stories like this move, I write about them in an email each morning.
One free email each morning, around 8 AM US Eastern.
Check your email.
I have sent you a confirmation link. One click and you are in.
Not in your inbox? Look in spam or promotions. If it's there, mark it "Not spam" so future editions land in your inbox.
The full picture
Perplexity launched Frontier Research on November 16, 2026, with the stated goal of creating "a system that can perform autonomous research". The division is backed by more than 10,000 GPUs, and work will be released openly and tested in Perplexity products, with regular publication of papers, models, data, and benchmarks. Focus areas include multi-agent orchestration, reinforcement learning, continual learning, systems, and pre-training. The company is hiring researchers and engineers in San Francisco, Palo Alto, New York, and London.
On November 17 and 18, 2026, Perplexity rolled out a Multiagent feature as a new effort level in its Computer and teammates products. In this mode, a lead agent splits tasks across a team of agents using different frontier models; agents share findings with one another during the process, and the lead agent combines all outputs into one answer. The feature is available to Max and Enterprise Max web users.
Before the Frontier Research launch, Perplexity published findings from red-teaming SPACE, its VM sandbox for computer agents. Testing covered 12 models on two tasks: escaping a VM to steal a host credential, and bypassing SPACE's network policy to reach a blocked URL. The work found that standard Capture the Flag setups used in red-teaming can steer agents away from the security boundary being tested, and that security-specialized models do not guarantee better results in agentic security boundary tests.
How it developed
Perplexity announced on November 16, 2026 a new research division called Frontier Research, with a stated goal of building "a system that can perform autonomous research".
The division is backed by more than 10,000 GPUs and will focus on multi-agent orchestration, reinforcement learning, continual learning, systems, and pre-training. Perplexity said work will be released openly, with regular publication of papers, models, data, and benchmarks, and tested in Perplexity products. The company is actively hiring researchers and engineers in San Francisco, Palo Alto, New York, and London.
Perplexity published red-teaming findings for SPACE VM sandbox, testing 12 models on VM escape and network policy bypass tasks
Sources
- Perplexity and NVIDIA Portable Computer launch
- DeepMind SIMA 2 and EVE Online partnership
- Perplexity's SPACE agent sandbox
- GPT-Rosalind agents and PerturbAI's brain atlas
BackgroundAI agents: models that take actions
Get the daily email
When stories like this move, I write about them in an email each morning.
One free email each morning, around 8 AM US Eastern.
Check your email.
I have sent you a confirmation link. One click and you are in.
Not in your inbox? Look in spam or promotions. If it's there, mark it "Not spam" so future editions land in your inbox.