Hey amigos, Perplexity just dropped some fuego updates in late February 2026, like on the 13th, 27th, making their AI smarter, more memoria-rich, and ready to handle big workflows. Think upgraded research that crushes benchmarks, a memory engine that remembers 95% of key stuff, and Perplexity Computer juggling 19 models for end-to-end projects. Cariños, whether youre coding, marketing, or analysing data, these changes mean less hassle, more results – lets dive in, mi gente.
—
✅ Upgraded Deep Research with Opus 4.6
Perplexity boosted their Deep Research tool to top external benchmarks like Google DeepMind Deep Search QA and Scale AI Research Rubric, using Opus 4.6 for Max and Pro users, with better memory and Model Council context. It pairs top models with their search engine for accurate, reliable deep dives.
Developers can analyse GitHub commit histories faster; marketers generate campaign briefs grounded in fresh data; analysts auto-summarise call transcripts from earnings calls. I used it last week to pull Shopify inventory trends – saved hours, amor.[1][2]
—
✅ Improved Memory Engine
Now creates richer memories and recalls details in 95% of cases, up from 77%, making half as many but way better ones. Rolled out to Max and Pro.
Ask it to remember book recs or past chats. Content writers get consistent tone across threads; researchers track evolving project notes; devs recall bug fixes from weeks ago. Dictated this bit voice-to-text style – it remembered my last query on Kimi K2.5 without prompting. Quality over quantity, si?[1][2]
—
✅ New Models: GPT-5.2 and Seedream 4.5
GPT-5.2 shines on coding like SWE-Bench Pro for software engineering and reasoning; Seedream 4.5 from Bytedance crafts cinematic images with better instruction following. Available to Pro and Max via model selector.
Devs tackle complex code; marketers whip up punchy ad visuals; analysts visualise data patterns. Tried Seedream for a client moodboard – spatial smarts nailed the layout, que nota.[1]
—
✅ Perplexity Computer
Launches as a multi-model system orchestrating 19 AI models for long-running workflows: research, code, design, deploy. Breaks tasks into subagents, integrates APIs, runs in cloud for Max subscribers at $200/month.
Marketers manage full campaigns end-to-end; devs sync inventory with Shopify via agents; teams build sites from financial data analysis. Practical for autonomous ops, like competitor onboarding flows.[3][4][5][8]
—
✅ Perplexity on Samsung Galaxy S26
Integrated deep into Galaxy S26, powering Bixby and Samsung Internet with real-time search and Comet-style agentic browsing. First phone with Perplexity APIs at platform level.[3]
On-the-go analysts query stock graphs; marketers check trends via voice; everyday users get grounded answers. Imagine Bixby pulling live Laver Cup scores – handy for sports fans hustling.[3]
—
✅ Upgraded Voice Mode in Comet
Voice now on GPT Realtime 1.5 for 25% more reliable chats, better expressiveness across tabs and screens. For all on desktop and Android.[3]
Sales teams talk through dashboards; researchers dictate notes on web pages; devs walk GitHub histories hands-free. Chatted with it over a messy tab setup yesterday – context stuck, no repeats.[3]
—
✅ State-of-the-Art Embedding Models
Released pplx-embed-v1 and pplx-embed-context-v1 (0.6B and 4B scales), topping Google and Alibaba benchmarks. For multilingual RAG and web-scale retrieval.[3]
Devs index docs in English, Japanese, German for queries; analysts scale billion-page searches affordably. Perfect for global campaign briefs, mija.[3]
Hey, thats the fresh batch – go play with Perplexity at https://www.perplexity.ai/, share your wins or feedback, and subscribe to their changelog for more. Cook loud, query proud!
