OpenAI: GPT-5 Will Be Offered Free to Users for Coding and Writing

In the live event this early morning, OpenAI officially launched the highly anticipated GPT-5 model, claiming it to be their most impressive model to date, showcasing top-tier capabilities in multiple fields such as programming, mathematics, writing, health, and visual perception.

Sam Altman, CEO of OpenAI, stated at the launch that GPT-5 is a major upgrade of the company's previous AI models, making interactions with it feel like talking to experts in various fields.

It is reported that GPT-5 will be gradually available to free users and paid users starting this Thursday. Paid users will have higher usage limits, while Pro subscribers can access GPT-5 unlimitedly and experience the powerful features of GPT-5Pro. Plus users can set it as the default model for daily questions, using it far more than free users.

GPT-5 is a highly unified system composed of three main parts: an intelligent and efficient model, a deep reasoning model (GPT-5thinking), and a real-time router. The intelligent and efficient model handles most questions, while the deep reasoning model specializes in complex problems. The real-time router intelligently selects the most suitable model for processing based on factors such as conversation type and complexity, continuously training and improving itself based on real signals. Once usage limits are reached, the streamlined versions of the models take over the remaining queries. In the future, OpenAI plans to integrate these features into a single model.

In benchmark tests, GPT-5 outperformed all previous models, not only answering questions faster but also providing more practical solutions, effectively reducing hallucinations, improving instruction following, and minimizing flattery in responses. Particularly in common ChatGPT application scenarios such as writing, programming, and health consultation, the performance of GPT-5 has seen significant improvements.

In the field of programming, GPT-5 is hailed as the strongest programming model to date, achieving significant progress in complex front-end generation and large codebase debugging. It can create aesthetically pleasing and responsive websites, applications, and games based on a single prompt. In creative expression and writing, GPT-5 can transform rough ideas into works with literary depth and rhythm, better handling writing tasks with unclear structures, and offering more assistance in daily writing. In the health sector, GPT-5 performs best in health-related questions, with significantly higher HealthBench scores compared to previous models. It can proactively identify potential issues and ask questions to provide more helpful answers, as well as offer more accurate and reliable responses based on user background, knowledge level, and location. However, OpenAI explicitly states that this model cannot replace medical professionals.

Additionally, GPT-5 has achieved new state-of-the-art levels in mathematics (94.6% score on AIME2025 without tools), real-world programming (74.9% on SWE-bench Verified, 88% on Aider Polyglot), multimodal understanding (84.2% on MMMU), and health (46.2% on HealthBench Hard).

Notably, GPT-5Pro, with its extended reasoning capabilities, also set a new state-of-the-art record on the GPQA test, achieving an impressive 88.4% score without tools. As the model designed for the most challenging and complex tasks by OpenAI, GPT-5Pro replaces the previous OpenAI o3-Pro. It has longer thinking time and uses scalable yet efficient parallel testing, providing the highest quality and most comprehensive answers.