We're integrating Deep Research as a native skill inside Computer. It now connects to the agent harness that powers Computer, with access to search as code generation, long running sandboxes, connectors, tools, and licensed data. Available now to Pro and Max subscribers.

Deep Research in Computer is built on our Search as Code architecture. The model writes code that assembles search itself, running thousands of retrieval steps in parallel, tailored to each question. It outperforms legacy Deep Research on every benchmark.

Learn more about Deep Research in Computer:

Perplexity is so underrated.

pplx deep research is actually best.. not with computer it will actually Reduce the hallucinations.. gonna try out today and share my new workflows.

What’s the deal with Comet search? Input a search term and it returns garbage phishing links. Type in Aetna. It returns. Aetna international makes no sense.

Deep Research inside the same agent harness as sandboxes, connectors and licensed data is the right architecture. Research that can't act is just a fancy tab.

Deep Research as a native skill in the agent harness is the right architecture decision. Research that can be triggered by other agents, not just by the user, changes what you can build on top. The harness layer is where the interesting orchestration actually happens.

I'll be taking advantage of this today! Perplexity keeps releasing new features and functionality that other AI services can't tough.

🔥🔥

To what extent is this your search infrastructure vs. you're now using Fable 5 rather than Opus 4.8?

@perplexity_ai This is huge, Deep Research + Computer agent = finally closing the “research → action” gap. No more copy-pasting outputs into tools or babysitting agents.

Interesting!

Generated search code raises one useful audit question: can an operator replay the exact retrieval path after a disputed answer?

every AI product announcement is just "we connected the thing to the other thing" with a different thumbnail

damn i have been using shallow research

the "long running sandboxes" part got me. last week i had openclaw spin one up to scrape and reconcile two years of messy invoices while i did other stuff, came back to a clean csv. didn't expect that to just work.

Deep Research gets much more useful when it can run inside the same harness that writes code, tests paths, and pulls licensed data.

@AravSrinivas Research is becoming less about finding information and more about turning information into action. That's a much bigger shift than most people realize.

Ah, you guys are awesome! Search is such a huge competitive advantage. Discovery is as important as any other part of the agentic workflow loop - and you're shipping pure magic right now.

agents getting agents now.

Deep Research is evolving from a feature into infrastructure. Search, code generation, long-running sandboxes, connectors, tools, and licensed data—all wired directly into the agent loop. The gap between "researching" and "doing" keeps getting smaller.

My coöputer credit has ended end of May and credit renewal date was shown as 09.06.2026. However, when that date arrived, it was suddenly changed to 09.07.2026 without any explanation. I have already submitted a mail about this but no reply till now. Please explain, why my credit didnot renewal and renewal date changed

agent loops skip the sandbox friction connectors run without prompt switching

This is a better Computer test than a one-off browser trick. When research, code gen, connectors, and sandboxes all touch the same task, I’d want the run to show which sources/tools it used before I trust the final answer.

I don’t get it. Why couldn’t legacy deep research also do that? What am I missing about why this integration is necessary for the performance delta?

Nice to see Perplexity actually shipping useful integrations instead of just talking about them

L'IA ne va pas remplacer les pros, ce sont les pros qui maîtrisent l'IA qui vont remplacer les autres. Prends une longueur d'avance dès aujourd'hui : 💼 #FuturDuTravail #IA #Prompting

@perplexity_ai has had 12 days to fix their sign in issues of my enterprise pro account and still NO HUMAN CONTACT and NO ETA to be fixed! What a joke!!!!!

Deep Research with live sandboxes and licensed data access is a different beast entirely. Research agents just got a serious upgrade.

search as code in parallel sandboxes is interesting. but agents running 50 retrieval steps cost money. when does accuracy gain beat the bill?

Congrats @AravSrinivas! Will set up @perplexity_ai computer on my lappy. 🫠🫠🫠👏👏👏👏👏

Benchmarks are jumping hard (e.g. BrowseComp 40.7% → 83.8%). This feels like a big step for agentic research. Perplexity just leveled up Deep Research by embedding it natively into Computer, routing subtasks across 20+ frontier models for better accuracy, reports, and dashboards. Great breakdown here:

finding papers is solved. the hard part: telling apart what actually works from what only works in the lab. that's what builders really need.