Model: Perplexity · Multi-Model
On 24 August 2026, Perplexity widened the set of models behind its answers — giving every Pro and Max user access to xAI’s Grok 4.6, making GPT-5.6 Terra and Luna the engines behind its agents and automations, and adding more frontier models to the mix. Here’s what a multi-model answer engine means for whether AI names your brand.
- On 24 August 2026, Perplexity expanded the models behind its answers.
- Grok 4.6 (xAI) is now available to all Pro and Max subscribers in Perplexity and Perplexity Computer.
- GPT-5.6 Terra is the new default for Computer subagents; GPT-5.6 Luna powers scheduled automations.
- The wider roster keeps growing, so the answer you see on Perplexity now depends heavily on which model is doing the work.
- For AI visibility, tracking has to run across models, not one engine — the same brand can be named by one and skipped by another.
What Perplexity actually changed
On 24 August 2026, Perplexity shipped a release it summed up as “Computer in Email, GPT-5.6 Terra and Luna, and Grok 4.6.” For anyone who cares about how AI answers questions, the headline is the model line-up. Grok 4.6, xAI’s latest model, is now available to every Pro and Max subscriber inside Perplexity and Perplexity Computer. GPT-5.6 Terra becomes the default model for Computer subagents and an option for orchestrating them, while GPT-5.6 Luna takes over as the primary model for scheduled automations. The release also extends Perplexity’s wider roster of frontier models and adds new agent and search tools for developers.
Alongside the models, Perplexity added Computer in Email — you can forward or send an email to start a Computer session, and it replies in the same thread using your own connectors, permissions and memory. Taken together, this is less a single feature and more a statement of direction: Perplexity is positioning itself as a multi-model answer engine and agent platform, not a single-model chatbot. For brands, the important shift is that the engine choosing what to say is no longer one fixed model.
Why a multi-model engine matters
When one model powered an answer engine, checking your visibility there meant checking one thing. A multi-model engine breaks that assumption. Perplexity now routes different work to different models — a fast model for a quick search, a heavier model for deep research, a specific model for a subagent or an automation. Each of those models was trained differently, weighs sources differently and phrases recommendations differently, which is the same reason the logic of how LLMs decide which brands to cite plays out unevenly from one engine to the next.
That means the “Perplexity answer” to your buying question is really several possible answers, depending on which model runs and which mode the user is in. A brand that a heavier model names confidently can be left out by a lighter one, and the reverse. We saw a smaller version of this whenever a model-picker change reshuffled results; Perplexity has now made model choice a core part of the product rather than a hidden setting.
What Grok 4.6 brings to the mix
Grok 4.6 is the most notable addition because it is a new model family on Perplexity, not an upgrade of one already there. Perplexity positions it for long-running, multi-step work — the kind of agentic research where a model plans, browses and cross-checks before answering — and highlights that it delivers that capability at a much lower cost than flagship models. Cheaper capable models get used more, so a model like this can quickly account for a real share of answers rather than sitting unused in the model selector.
For visibility, a capable, agentic model is the same double-edged story we described with Claude Opus 5 and Gemini 3.7 Flash: it researches and compares more thoroughly, which rewards brands with clear, consistent public evidence and filters out thin or contradictory ones. The wrinkle here is provenance — Grok is xAI’s model running inside Perplexity, so its instincts about which sources to trust are its own, not Perplexity’s. Being named well on one model on the platform tells you nothing certain about the others.
How model routing shapes answers
The practical effect of routing is variability. Ask Perplexity the same question in a normal search, in deep research mode, and via an automation, and you may be handed to three different models — with three different shortlists of brands. None of that is visible to the person asking; they simply see “Perplexity said.” So the visibility you had last month under one model is not guaranteed this month under another, exactly as it is not guaranteed after any fast-answer change to how responses are generated.
This is why a single spot-check is a weak signal on a multi-model engine. What matters is coverage across the models and modes your customers actually use, watched over time. The good news is that the underlying discipline does not change per model: the brands that surface consistently are the ones whose facts are easy for any capable model to read, verify and repeat — the core of building a GEO strategy.
What the Perplexity multi-model update changes
Perplexity announced its multi-model update on 24 August 2026.
What email agents and automations do
Computer in Email, subagents and scheduled automations all point the same way: more of the journey runs through agents rather than a person typing a question. When GPT-5.6 Terra drives a subagent or GPT-5.6 Luna runs a nightly automation, a model is doing the reading, filtering and shortlisting on someone’s behalf — often a lightweight model that visits your page, extracts a fact and decides whether you clear a filter before any human sees the result. We have tracked that shift from ChatGPT Sites to Perplexity Brain, and this release deepens it.
The test for your brand is machine-readability. If an agent can parse who you serve, what you solve and the constraints you suit in seconds, you clear more of those automated gates; if that evidence is buried in images or inconsistent across pages, you get skipped silently. It is the same bar we set in is your brand ready for AI search? — only now it is applied by more models, more often, inside more of the workflow.
Does being on more models help you?
It is tempting to read “more models on Perplexity” as more chances to appear, and that is partly true — more models generating answers is more surfaces where a well-evidenced brand can be named. But it is not a lever you can pull. Being available on a platform, or being a model people can select, does not make any model recommend you; that still comes down to whether your public evidence is the clearest match for the question, the point we keep returning to from GPT-5.5 Instant recommendations to preferred sources for AI Overviews.
The realistic takeaway is that multi-model raises the variance. Your brand can win on one model and lose on another for the same query, so an average across the engine matters more than any single lucky result. That is an argument for measurement, not for chasing one model — and for the kind of clean, consistent content that travels well across all of them, which is also what Google’s own llms.txt guidance rewards.
How to check your Perplexity visibility
Because Perplexity now spans several models, one check on one model is no longer a fair read. Here is a short way to re-baseline — today is 26 August 2026:
- List the models and modes. Note which models you can select and which modes — quick search, deep research, Computer — your customers are likely to use.
- Test each in turn. Run your key buying questions on each available model, including Grok 4.6, and record which brands Perplexity names.
- Watch the sources. Note which pages each model cites, and whether your own content is among them.
- Compare model to model. Flag any brand that appears on one model but vanishes on another — that gap is your real exposure.
- Repeat in automations. If your audience uses scheduled or agentic workflows, check what those return, since they may run a different default model.
- Track the movement. Models and defaults change often, so treat each run as a snapshot and watch the trend over time.
How reconnAI tracks Perplexity answers
reconnAI monitors how the leading AI models — ChatGPT, Claude, Gemini, Perplexity, Copilot and Google AI Overview — answer questions across regions, and re-baselines whenever a platform changes the models behind its answers. Perplexity turning into a multi-model engine is exactly that kind of moment: the thing generating your answer is now a set of models, and each can name or skip you differently.
It joins a busy run of shifts we have followed lately — from Perplexity Deep Research and Claude Opus 4.8 to Sign in with ChatGPT and expanded custom instructions. If you want to see how Perplexity represents your brand across its new line-up, get in touch with our team or explore how AI visibility tracking works. If your category can be reached commercially, our overview of ChatGPT advertising covers where paid AI placement is heading.
Perplexity models: your questions answered
What did Perplexity change on 24 August 2026? It expanded the models behind its answers — adding xAI’s Grok 4.6 for all Pro and Max users, making GPT-5.6 Terra the default for Computer subagents and GPT-5.6 Luna the model for scheduled automations, and adding Computer in Email plus new developer tools.
Does being on Grok 4.6 or GPT-5.6 change my ranking? Not directly. Which model is available does not decide whether it recommends you; that still depends on whether your public evidence is the clearest, most consistent match for the question. A capable model simply reads and compares more thoroughly.
Why does a multi-model engine matter for AI visibility? Because different models cite and recommend differently. The same Perplexity question can produce different brand shortlists depending on which model and mode runs, so your visibility has to be measured across models, not on one.
Is Grok 4.6 a Perplexity model? No. Grok 4.6 is built by xAI and made available inside Perplexity for Pro and Max subscribers, while GPT-5.6 Terra and Luna are OpenAI models. Perplexity is acting as a router across several providers’ models.
About reconnAI
reconnAI tracks how the major AI models represent topics and sources across ChatGPT, Claude, Gemini, Perplexity, Copilot and Google AI Overview — across multiple regions. We monitor how those models answer and how they change over time, so you can stay ahead of shifts in the AI landscape.
Last updated: 26 August 2026.