llm architecture

2 articles

GPT-5 Router Explained: Selecting Thinking vs. Fast Models

Learn how OpenAI's GPT-5 router works. This article explains its unified architecture and the criteria it uses to route queries to fast or thinking sub-models.

10/13/2025• 32 min read

gpt-5 routeropenailarge language models

Perplexity's LLM: A Technical Deep Dive on Sonar & PPLX

Does Perplexity have its own LLM? We explain its proprietary Sonar & PPLX models, its hybrid tech stack using GPT-4, and its unique answer engine architecture.

10/12/2025• 42 min read

perplexity ailarge language modelpplx