<p>OpenAI has officially announced GPT-5, their most advanced language model to date, featuring significant improvements in reasoning capabilities, context length, and pricing structure. Released on August 7, 2025, GPT-5 introduces three model variants designed to meet different performance and cost requirements.</p> <p>The flagship GPT-5 model offers a massive 272,000 token input context window with 128,000 tokens for output, representing a substantial increase over previous generations. The model features four distinct reasoning levels and achieves dramatically reduced hallucination rates compared to GPT-4.</p> <p>OpenAI introduces a competitive pricing structure with GPT-5 at $1.25 per million input tokens and $10 per million output tokens. GPT-5 Mini offers similar capabilities at $0.25/$2 per million tokens, while GPT-5 Nano provides basic functionality at $0.05/$0.40 per million tokens.</p> <p>A standout feature is the 90% token caching discount, which significantly reduces costs for repeated queries with similar context. The full GPT-5 model maintains a knowledge cutoff of September 30, 2024, while Mini and Nano variants use data up to May 30, 2024.</p> <p>"GPT-5 represents our most significant leap forward in AI reasoning and reliability," OpenAI stated in their announcement. "The model's enhanced context understanding and reduced hallucinations make it particularly suited for complex professional applications."</p> <p>The release positions OpenAI to compete directly with other frontier models while offering developers more granular options for balancing performance and cost across different use cases.</p>