Best
The automatic option in Search. It picks a suitable model for each question. Not a separate model.
Perplexity lets paid subscribers choose between its own Sonar model and leading models from OpenAI, Anthropic, Google, xAI and others, all with Perplexity’s search and citations. Here’s which models each plan includes, how Best, Thinking and Model Council work, and a tool to help you pick.
In short
Free users get Best, which picks a model automatically. Pro adds a choice of eight models, including Sonar 2, GPT-5.6 Terra, Claude Sonnet 5, Gemini 3.7 Flash and Grok 4.6. Max adds GPT-5.6 Sol and Claude Opus 5, plus Model Council to consult several models at once. Turn on Thinking for hard, multi-step questions. Deep Research picks its own models.
Availability
Search models listed in Perplexity’s help center as of 4 September 2026. Enterprise Pro and Enterprise Max get the full list.
| Model | Maker | Free | Pro | Max |
|---|---|---|---|---|
| Sonar 2Perplexity’s own | Perplexity | Best only | ✓ | ✓ |
| GPT-5.6 Terra | OpenAI | Best only | ✓ | ✓ |
| Gemini 3.7 Flash | Best only | ✓ | ✓ | |
| Claude Sonnet 5 | Anthropic | Best only | ✓ | ✓ |
| Kimi K3 | Moonshot AI | Best only | ✓ | ✓ |
| GLM 5.3 | Z.ai | Best only | ✓ | ✓ |
| Grok 4.6 | xAI | Best only | ✓ | ✓ |
| Nemotron 3 Ultra | NVIDIA | Best only | ✓ | ✓ |
| GPT-5.6 SolMax only | OpenAI | Best only | — | ✓ |
| Claude Opus 5Max only | Anthropic | Best only | — | ✓ |
Showing all 10 models
Models change often, and your plan, region or a staged rollout can affect what you see. The model selector in your account is always the source of truth. Source: Perplexity Help Center.
Options
The automatic option in Search. It picks a suitable model for each question. Not a separate model.
Gives the model more time to reason. Use it for complex reasoning, planning, coding or multi-step analysis; answers take longer.
Consult several models at once for research, complex decisions or verification, with a combined answer.
Chooses its own models; you can’t pick one. Deep Research guide.
Comet and Perplexity Computer have their own model selectors and rules; Computer tasks use credits. Computer guide.
Choose your task and plan. Suggestions follow Perplexity’s own guidance, plus common user preferences.
Step by step
Pro and Max only. Labels and positions can differ between the website and apps.
Families
General descriptions, not benchmark scores. In Perplexity, every model works with Perplexity’s search and citations, so answers differ from the makers’ own apps.
Perplexity’s own model, built for fast, grounded answers with citations. A good default for everyday search.
General-purpose models; Sol is OpenAI’s higher-end reasoning option on Max.
Often chosen for writing and careful analysis; Opus 5 is Anthropic’s top tier, on Max.
A fast Google model; handy for quick answers and mixed text-and-image questions.
xAI’s model, described by some reviewers as suited to longer, multi-step work.
Alternative advanced models worth trying when you want another perspective.
Tips
FAQ
As of 4 September 2026, Perplexity’s help center lists Sonar 2, GPT-5.6 Terra, Gemini 3.7 Flash, Claude Sonnet 5, Kimi K3, GLM 5.3, Grok 4.6 and Nemotron 3 Ultra for Pro and Max, with GPT-5.6 Sol and Claude Opus 5 added on Max. The list changes as providers release new models.
No. Free users get Best, which picks a suitable model automatically, plus a limited number of Pro Searches. Choosing a specific model needs Pro or Max.
Best isn’t a separate model or mode. It’s the automatic option in Search that picks a suitable model for each question, so you don’t have to.
Thinking gives a model more time to reason before answering. It helps with complex, multi-step questions, planning and coding, but answers take longer. It’s a setting on supported models, not a separate model.
Model Council lets you consult several models at once for research, complex decisions or verification, and combines their answers. It’s available on Max and Enterprise Max, on the web.
No. Deep Research selects its models automatically.
Not exactly. Perplexity runs third-party models with its own search, citations, tools and safety systems, so answers can differ from the makers’ own apps.
Your plan, your region, a staged rollout or (for Enterprise) an administrator setting can limit which models you see. The model selector in your account is always the source of truth.
No. The API is billed separately from Free, Pro and Max subscriptions. See our API guide.
Try it
On Pro or Max, open the model selector and compare. On Free, Best picks for you.