What Model Does Perplexity Use? 2026 List: Sonar 2, GPT-5.6, Claude, Gemini
Updated: 1 day ago
Perplexity's 2026 model lineup is split across Search and Computer, and access differs between Pro and Max rather than every paid user seeing the same list.
The current model map includes Perplexity's own Sonar family plus models from OpenAI, Anthropic, Google, xAI, Moonshot AI, Z.ai, and NVIDIA. The selector can also differ by region, rollout, and enterprise policy.
··········
SEARCH AND COMPUTER DO NOT SHARE ONE MODEL PICKER
Search exposes models optimized for cited answers and reasoning, while Computer can use orchestrator and subagent models for longer workflows. Perplexity's August 2026 updates added GPT-5.6 Terra and Luna to Computer roles and expanded Grok 4.6 availability.
........
Model | Provider | Typical plan access | Role |
|---|---|---|---|
Sonar 2 | Perplexity | Pro and Max | Perplexity-native search model |
GPT-5.6 Terra | OpenAI | Pro and Max | General advanced reasoning |
GPT-5.6 Sol | OpenAI | Max in Search; also used in Computer where available | Higher-end reasoning |
Gemini 3.1 Pro | Pro and Max | Reasoning and multimodal work | |
Claude Sonnet 5 | Anthropic | Pro and Max | General advanced model |
Claude Opus 5 | Anthropic | Max | Highest Anthropic tier |
Grok 4.6 | xAI | Pro and Max | Long-running and multi-step work |
Kimi K3 / GLM 5.2 / Nemotron 3 Ultra | Multiple | Varies by plan and workspace | Alternative advanced reasoning models |
........
··········
PRO COVERS MOST ADVANCED SEARCH MODELS
Pro includes advanced model selection rather than forcing the automatic default on every query. Current official plan documentation lists models such as Sonar 2, GPT-5.6 Terra, Gemini 3.1 Pro, Claude Sonnet 5, Grok, and additional third-party models depending on rollout.
··········
MAX UNLOCKS THE HIGHEST-END OPTIONS
Max adds the highest access level and models that are not normally part of Pro Search, including GPT-5.6 Sol and Claude Opus 5 in the current model matrix. Max also raises limits across Research, Computer, and creation tools.
··········
GROK 4.6 REPLACED THE OLDER 4.5 ASSUMPTION
Perplexity's August 24 changelog states that Grok 4.6 is available to all Pro and Max subscribers in Perplexity and Perplexity Computer. Older help-center tables may still show Grok 4.5, so the changelog is the more recent product signal.
··········
BEST MODE AND RESEARCH MODE CAN CHOOSE FOR YOU
[[ADPLUS_300x250_1]]
Not every workflow requires manual model selection. Best mode can route the request automatically, while Research selects an appropriate combination of models for autonomous investigation.
··········
ENTERPRISE MODEL LISTS CAN BE SHORTER
Enterprise administrators can disable specific model providers and apply stricter retention policies. A model shown in the public plan matrix therefore may not appear inside every managed workspace.
··········
HOW TO CHOOSE A MODEL IN PRACTICE
Choose the model only when its behavior materially changes the task. For routine cited search, automatic routing is usually sufficient; for complex reasoning or long-form synthesis, selecting a stronger model can be useful; for agentic work, the Computer model matrix matters more than the Search picker.
··········
FOLLOW US FOR MORE
··········
DATA STUDIOS
··········
