Perplexity Upgrades Pro Subscription with GPT-5.2 and Claude 4.6 Reasoning Models

Perplexity has updated its Pro subscription to include the latest generation of frontier AI models, offering users a unified interface to access proprietary tools from OpenAI, Anthropic, Google, and xAI. This early 2026 update highlights the arrival of GPT-5.2 and Claude 4.6. Users can now toggle between distinct personalities and technical strengths depending on whether they need complex coding or creative writing. A major addition to the platform is the Reasoning toggle. This feature allows models such as GPT-5.2 and Grok 4.1 to engage in advanced logical processing for complex queries. By activating this mode, the AI performs internal logic checks before generating a response, which helps in reducing factual errors and hallucinations according to the latest documentation.

The Pro Search function remains a cornerstone of the service. It utilizes these advanced models to perform multi-step internet investigations, providing three times the number of sources compared to a standard search. This deep research capability is designed for users who require thorough documentation and varied perspectives on difficult topics.

Google’s Gemini 3.1 Pro and Moonshot AI’s Kimi K2.5 take a different approach by keeping reasoning permanently enabled. Gemini 3.1 Pro excels at multimodal tasks such as analyzing charts or video, while Kimi K2.5 focuses on privacy first logic and step by step technical explanations. These models ensure a high level of analytical depth for every search without requiring manual adjustments.

For software developers, Claude Sonnet 4.6 has become the preferred choice due to its efficient coding and technical reasoning. While most models are accessible through the standard Pro tier, the flagship Claude 4.6 Opus is restricted to Max subscribers. This tiering reflects the high compute costs associated with the most demanding agentic tasks.

The platform manages usage through a rolling 24 hour credit system rather than a fixed daily reset. Users can also utilize the Best mode, an automated router that selects the ideal model for a query based on its specific requirements. This allows subscribers to benefit from the unique strengths of various LLMs, which process data through building blocks known as tokens, without needing to be experts in AI architecture.