Perplexity

Perplexity AI specializes in search-grounded language models. Unlike conventional AI models that generate responses from training data, Perplexity’s Sonar models search the web as part of every response — meaning responses are always grounded in current, cited sources.

---

​

Available Models

---

​

What Makes Perplexity Different

Most AI models generate responses from training data and only optionally search the web (when you activate the Search pill). Perplexity’s Sonar models work differently:

• Always web-connected: Every Sonar response involves a web search, regardless of whether you’ve activated the Search pill
• Built-in citations: Responses include source citations automatically — you can see exactly where each piece of information came from
• Current information by default: You don’t need to ask for “current” or “recent” information — Sonar is always looking at live web content

This makes Sonar models uniquely suited for research, fact-checking, and any query where having the most up-to-date information matters.

---

​

Strengths

Research questions: Sonar Pro is excellent for multi-faceted research questions that benefit from synthesizing multiple recent sources. Current events: For news, market conditions, product releases, and any time-sensitive topic, Sonar’s real-time web grounding is a significant advantage. Fact-checking: Sonar’s citation system makes it easy to verify claims — every assertion links back to a source you can inspect. Cited responses: For professional or academic contexts where source attribution matters, Sonar provides this automatically.

---

​

Best Use Cases

---

​

When to Use Perplexity vs. Web Search

You can add web search to any model by activating the Search pill in ZeroTwo. How is that different from using Sonar directly? Use Perplexity Sonar when: you want comprehensive, always-cited responses and don’t want to think about activating search each time. Use Web Search + another model when: you prefer a specific model’s response style but want web augmentation for specific queries.

---