OpenAI officially launches its anticipated GPT-5 model for both free & paid ChatGPT users

OpenAI has officially released GPT-5, its highly anticipated and most advanced large language model, granting access to all ChatGPT users, including those on the free tier. The update brings substantial improvements in reasoning, coding, memory, and multimodal understanding, resulting in more accurate and humanlike responses. GPT-5 also sets new marks on core AI benchmarks, such as achieving 94.6% on AIME 2025 for math, up to 88% on key coding challenges, and 84.2% on multimodal tests.

Three main versions are available: GPT-5 (standard), GPT-5 Mini (active during rate limits), and GPT-5 Nano (low-cost), each offering a 256K context window and 128K output token maximum. For the ChatGPT Pro-tier, GPT-5 Pro delivers enhanced reasoning, top performance on complex tasks, and unlimited usage at $200 a month. GPT-5 introduces chain-of-thought reasoning, reduces hallucinations by 80% compared to the o3 model, supports “safe completions” for sensitive prompts, and offers stronger memory feature for recalling user details across sessions.

The advanced Voice Mode has also seen improvements, now adapting to emotional tone, while new preset personalities (Cynic, Robot, Listener, Nerd) enable more personalized chats. Developers can access GPT-5 via the OpenAI API, Microsoft Azure, and GitHub Copilot (starting with Copilot Enterprise and Pro+), including integration in GitHub Chat, Visual Studio Code, and GitHub Mobile, with API pricing starting at $1.25 per million input tokens. Despite the significant advancements in the new models, it likely still falls short of the long-rumored AGI, as Sam Altman himself acknowledged by noting that it still lacks continuous self-learning, even if its capabilities represent a major leap over previous versions.