Perplexity

A 0.6B parameter bidirectional text encoder that detects and masks personally identifiable information in conversational data. It performs token level classification across 9 PII categories using a BIOES tagging scheme…

• 🔒 Data anonymization
• 🔒 Data protection
• 🔒 Data leak prevention
• 🔍 Data classification New

Pplx pplx-embed is a family of text embedding models built for real-world, web-scale retrieval.

• 🔍 Information retrieval
• 📂 Content categorization
• 🔍 Conceptual search

A 4B-parameter text embedding model built on diffusion-pretrained Qwen3. Produces 2560-dimensional dense embeddings with INT8 and binary quantization, 32K context window, and Matryoshka Representation Learning (MRL) sup…

• 🔗 Embeds
• 🔍 Information retrieval
• 🔍 Conceptual search

A 0.6B parameter multilingual text embedding model built on a diffusion-pretrained Qwen3 backbone. Produces 1024-dimensional unnormalized embeddings with int8 and binary quantization support, a 32K token context window,…

• 🔍 Information retrieval
• 🗄️ Vector databases
• 📚 Non-interactive language modeling

A 0.6B-parameter contextual text embedding model for RAG pipelines. Unlike standard embedding models, it takes full document chunks together so each chunk's embedding reflects surrounding context. Produces 1024-dimensio…

• 🔍 Information retrieval
• 📚 Knowledge bases
• 🔍 Document search
• 🔢 Vector indexes

A 4B-parameter contextual text embedding model built on diffusion-pretrained Qwen3, designed for RAG pipelines where document chunks benefit from surrounding context. Produces 2560-dimensional INT8/BINARY-quantized embe…

• 🔍 Information retrieval
• 🔍 Document search
• 🗄️ Vector databases
• 🔢 Vector indexes

Perplexity’s open-source, post-trained variant of DeepSeek-R1 that removes CCP-aligned censorship to deliver unbiased, factual answers while preserving R1’s reasoning quality. It’s available as weights on Hugging Face a…

• ❓ Answers

Perplexity’s premium reasoning assistant. It goes beyond quick Q&A by systematically searching across trusted sources, synthesizing information into a structured report, and exposing citations so you can verify or dive…

• 🔍 Research
• 💑 Virtual girlfriend
• 💪 Motivational coach
• 📝 Essay outlines

Perplexity’s balanced reasoning model, sitting between the smaller Sonar options and the top-end Sonar Large. It’s tuned for strong multi-step reasoning, accurate retrieval grounding, and code assistance, with a practic…

• 💻 Coding
• 📊 Data analysis
• 📝 Writing
• 🔍 Research

Perplexity’s family of reasoning models built on the R1 post-training architecture. They’re optimized for retrieval-augmented generation (RAG), accurate instruction following, and structured outputs, with multiple sizes…

• 📝 Writing
• 💻 Coding
• 🤖 ChatGPT
• 🔍 Research

Sonar Perplexity Sonar Large 32K is the top-tier Sonar reasoning model, built on Perplexity’s R1 post-trained architecture. It supports a 32,000-token context window, excels at deep reasoning and coding, and is tuned fo…

Sonar Advanced reasoning model combining multi-step Chain-of-Thought processing with real-time web search integration. Features a 128K context window. Successor to the deprecated sonar-reasoning model, built for complex…

• 🔍 Research
• 📊 Data analysis
• 💬 Chatting
• 📝 Text generation New