Perplexity AI All Models Available: list, categories, usage, etc

Perplexity AI provides access to a wide set of models across its app and API, combining its in-house Sonar family with leading frontier models such as GPT-5, Claude 4, Gemini 2.5, and Grok. In late 2025, the service is structured into search, reasoning, and research categories, with distinctions between Pro and Max subscriptions and a dedicated Sonar lineup for API use. Understanding the differences between each surface and plan is essential to choose the right model for speed, reasoning, or long-form research.

·····

.....

Pro and Max plan model options in the app.

Perplexity’s app allows users on Pro and Max tiers to select models per thread. In Pro Search, the available models are Sonar, GPT-5, Claude 4.0 Sonnet, and Gemini 2.5 Pro. These options support deeper multi-source answers and manual control over the model used for retrieval.

In Reasoning Search, Pro users can choose between o3, Claude 4.0 Sonnet Thinking, and Grok 4. The Max plan adds access to o3-pro, Claude 4.1 Opus Thinking, and GPT-5 Thinking, which are reserved exclusively for Max subscribers. This tier provides early and priority access to frontier models and reasoning-optimized versions.

Two additional modes automate model selection. Best mode automatically chooses the most appropriate model for speed and accuracy. Research mode dynamically composes a mix of models to produce in-depth, source-rich reports. These modes are designed for users who prefer not to make manual selections.

·····

.....

The role of Sonar as Perplexity’s default model.

The default model for quick searches is Sonar, Perplexity’s in-house model family. Built on Llama 3.3 70B and fine-tuned for factual accuracy and readability, Sonar powers fast web-grounded answers. It provides the foundation for both free and Pro usage, ensuring consistency across sessions. Pro users can set Sonar as the default, while other models remain available for manual selection.

Sonar is distinct from base Llama checkpoints; it is optimized for Perplexity’s retrieval-augmented interface. This means that while it shares lineage with Llama, Sonar outputs are tuned for the factual, citation-linked style that characterizes Perplexity’s responses.

·····

.....

The API lineup and the Sonar family.

The Perplexity API does not expose GPT, Claude, or Gemini directly. Instead, it provides the Sonar family of models grouped by task:

• Search: sonar for lightweight, grounded search and sonar-pro for deeper retrieval with follow-ups.

• Reasoning: sonar-reasoning for real-time reasoning with search, and sonar-reasoning-pro, which is powered by DeepSeek-R1 and exposes visible reasoning content through the API.

• Research: sonar-deep-research, which produces long-form, source-dense reports. This model supports asynchronous jobs and includes a reasoning_effort parameter to adjust the depth of analysis.

Earlier identifiers such as llama-3.1-sonar-*-128k-online and pplx-* have been deprecated. Developers are expected to migrate to the new sonar* family, which unifies naming and functionality across tasks.

·····

.....

Model selection by mode in the app.

In the app, models are surfaced differently depending on the mode:

• Best mode: automatic model choice for quick answers.

• Pro Search: manual selection between Sonar, GPT-5, Claude 4.0 Sonnet, or Gemini 2.5 Pro.

• Reasoning Search: choice of o3, Claude 4.0 Sonnet Thinking, or Grok 4; with Max-only options o3-pro, Claude 4.1 Opus Thinking, and GPT-5 Thinking.

• Research mode: an automatic blend of models for comprehensive reporting.

This layered structure gives users control when needed, while still offering automation for broader tasks.

·····

.....

Summary table of models and availability.

Surface	Category	Models	Notes
Perplexity App (Pro)	Pro Search	Sonar, GPT-5, Claude 4.0 Sonnet, Gemini 2.5 Pro	Manual choice
	Reasoning	o3, Claude 4.0 Sonnet Thinking, Grok 4	Grok excluded for Enterprise
	Best	Auto	Default quick mode
	Research	Auto mix	Builds long reports
Perplexity App (Max)	Reasoning extras	o3-pro, Claude 4.1 Opus Thinking, GPT-5 Thinking	Max-only access
Perplexity API	Search	sonar, sonar-pro	Grounded web search
	Reasoning	sonar-reasoning, sonar-reasoning-pro	Pro version powered by DeepSeek-R1
	Research	sonar-deep-research	Async, reasoning_effort knob

The table illustrates how availability differs between consumer subscriptions and the developer API, with Max unlocking additional reasoning models and the API consolidating everything under Sonar.

·····

.....

Plan restrictions and naming considerations.

Not all models are available on all plans. Max subscribers enjoy access to o3-pro, Opus 4.1 Thinking, and GPT-5 Thinking, while Pro subscribers are limited to core Pro Search and Reasoning models. Enterprise customers are excluded from Grok 4 access due to licensing restrictions.

Naming conventions have also shifted. The Sonar family replaces earlier llama-3.1-sonar and pplx identifiers. Developers must use the current Sonar endpoints for Search, Reasoning, and Research. This transition reflects Perplexity’s move to a unified model branding that separates consumer-facing model names from API designations.

·····

.....

Operational recommendations.

For quick, grounded answers with citations, users should rely on Sonar in either the app or API. For more demanding reasoning tasks, Reasoning Search in the app or the sonar-reasoning endpoints in the API provide better multi-step analysis. When comprehensive, source-rich reports are required, Research mode in the app or sonar-deep-research in the API is most effective.

Users who want direct access to frontier models such as GPT-5 or Claude Opus should subscribe to Max, while developers should align their applications with the Sonar family endpoints. Because Perplexity frequently updates model availability and deprecates older identifiers, it is best practice to confirm the latest model roster on official documentation before deploying.

.....

FOLLOW US FOR MORE.

DATA STUDIOS

.....[datastudios.org]