I was nursing a flat white in my Melbourne laneway spot last week, scrolling Perplexity’s changelog on my phone, when March 2026’s updates hit me like a sudden sketch inspiration. Perplexity just rolled out heftier tools for research, automation, and cross-platform magic, all dated within the last fortnight from their 27 March ship. Developers, marketers, analysts – grab your Pro or Max sub; these changes mean workflows that hum smoother, answers that dig deeper, and tasks that run themselves.
—
Comet iOS Launch and Computer Updates
Perplexity pushed Comet to iOS and big platforms on 27 March, beefing up Computer with inline editing, task controls, live credit tracking, and better handling of long chats. Deep Research now spits out presentations and structured outputs; Health and Finance got fresh tricks too.
For developers, tweak code snippets inline without app hopping, like refining a Shopify inventory sync script on the fly. Marketers draft campaign briefs in structured slides, auto-formatted for client pitches. Analysts track research credits live while auto-summarising call transcripts from Zoom recordings. Researchers build finance timelines that pull live data, no manual fiddling.
—
GPT-5.4 and GPT-5.4 Thinking Rollout
On 6 March, Perplexity made OpenAI’s GPT-5.4 models available to all Pro and Max users across web and mobile, with Thinking mode for step-by-step reasoning on tough problems. Gains in coding, reasoning, creative work.
Developers debug complex Node.js apps faster with chained reasoning. Content writers generate LinkedIn threads that flow naturally, step by reasoned step. Marketers brainstorm A/B test variants for Google Ads, each backed by logic trails. Analysts model revenue forecasts, breaking down assumptions one by one.
—
Upgraded Deep Research
Deep Research hit state-of-the-art benchmarks around early March, now on Opus 4.5 for Max and Pro users soon after. Pairs top models with Perplexity’s search and sandbox for accurate, reliable deep dives; creates presentations too.
Researchers compile competitor analyses from 50 sources into slide decks. Marketers pull market trends into reports for stakeholder decks. Developers document API integrations with benchmark-backed explanations. Analysts dissect earnings calls into key tables and visuals.
—
Custom Skills and Model Council
Shipped 6 March: Custom Skills automate repeat tasks, Model Council runs frontier models in parallel for consensus answers, plus Voice Mode and GPT-5.3-Codex for coding.
Set a skill to auto-schedule Notion tasks from email threads – saves analysts hours weekly. Developers trigger code reviews via voice, council cross-checks outputs. Marketers run parallel idea gen for email campaigns, pick the best consensus. Content writers dictate outlines, get polished drafts back.
—
Improved Memory Engine
Rolled out to Max and Pro post-6 March: richer memory creation and recall from past chats. Ask it to remember book recs or convo details.
Marketers reference prior campaign feedback without re-explaining. Developers recall yesterday’s API tweaks mid-session. Analysts build on last week’s data pulls seamlessly. Researchers string multi-day projects without context loss.
—
Pop over to Perplexity.ai, test these in your Pro trial, and drop feedback in their changelog comments. Subscribe for the next batch – my pencil’s already itching for what’s next.
