“Sonar” refers to multiple advanced AI offerings grouped under the same name, each optimized for different real-world needs. Perplexity’s Sonar is a search-and-reasoning LLM tuned to deliver instant, highly factual answers with cited sources. It emphasizes speed (very low latency), strong multi-step reasoning, real-time web search integration, and privacy-first hosting — making it ideal for professionals who need concise, trustworthy answers without digging through links. Developers can access it via Perplexity Pro or the Sonar API. Meta’s SONAR is a multilingual, multimodal sentence-embedding system that maps text and speech into a single fixed-size semantic space. It supports ~200 languages, powers fast semantic similarity search, and enables text-to-text and speech-to-text translation (including zero-shot transfers). It’s best for cross-lingual retrieval, large-scale embedding search, and translation pipelines where consistent, compact embeddings matter. Separately, Sonar-branded tools in the software-development ecosystem (SonarQube / SonarCloud) add generative-AI features for code quality and security — identifying vulnerabilities, suggesting fixes, and applying AI CodeFix recommendations to reduce bugs and outages from both human- and AI-written code. Practical benefits: use Perplexity Sonar when you want near-real-time, citation-backed answers and deep reasoning; use Meta SONAR when you need efficient, language-agnostic embeddings for multilingual search, retrieval, or speech+text translation; and use Sonar code tools to automate code review, vulnerability detection, and safe remediation. Limitations and access: Perplexity Sonar is a proprietary product available via subscription/API; Meta SONAR is a research-grade embedding service focused on cross-modal multilingual tasks; Sonar code tools are specialized for software quality workflows. Choosing the right Sonar depends on whether your priority is fast factual Q&A, scalable multilingual embeddings, or code quality automation.