sonar

Sonar is a new generation of intelligent inference model series launched by Perplexity AI, with its core positioning as the next-generation AI search and generation base for "real-time internet connection + efficient inference", deeply integrating web search capabilities to achieve "what is answered is what is searched".

• Native Internet Connectivity Architecture: No external RAG required—this model integrates a real-time search engine internally, automatically retrieving, verifying, and citing the latest authoritative information.
• Extremely Low Hallucination Rate: All generated content is based on trustworthy sources, with one-click access to cited links, ensuring factual accuracy.
• Multi-Language Support: Covers dozens of languages including Chinese, English, Japanese, German, French, and more, delivering natural and culturally appropriate outputs.
• Developer-Friendly API: Offers advanced features such as structured JSON output, streaming responses, and customizable search scopes (e.g., limited to academic sources, news, or official websites).

───────────────────────────────────────────────────────────────────

Core Capabilities

🔍 AI-Enhanced Search: Upgrades traditional keyword-based searches to natural language question answering—for example, “What’s the global market share ranking of AI chips in 2025?”

⚡ Second-Level Knowledge Integration: From query submission to returning a structured answer with citations takes only hundreds of milliseconds—far faster than manually checking multiple web pages.

🧠 Context-Aware Reasoning: Understands references and intentions across multi-turn conversations (e.g., in “What technology did it use?,” the pronoun “it” refers to the model mentioned earlier).

🧩 Agent-Ready: Can serve as the core engine for research agents, market analysts, or technical consultants, automatically completing the closed loop from information gathering → analysis → report generation.