OpenAI, an artificial intelligence research and deployment company, has introduced GPT-5, its most advanced AI model to date, now available to all ChatGPT users. The system integrates fast-response and deep-reasoning capabilities, enabling it to choose the optimal approach based on task complexity.
GPT-5 delivers improved performance in coding, writing, mathematics, health, and multimodal reasoning, surpassing previous benchmarks in accuracy and real-world utility. Paid tiers receive extended access, with Pro subscribers also gaining GPT-5 Pro for high-precision, complex queries. OpenAI reports significant reductions in factual errors and hallucinations, along with improved instruction adherence and less sycophantic behavior.
The model’s capabilities include enhanced design sensibility in software development, advanced literary composition, and more reliable health-related guidance, scoring higher on evaluations like HealthBench. It handles complex, multi-step requests more effectively and provides clearer communication when tasks are unfeasible. GPT-5 introduces a “safe completions” safety approach, offering nuanced responses in dual-use or ambiguous cases. GPT-5 Pro outperformed other versions in 67.8% of expert evaluations for economically valuable reasoning tasks. The rollout extends to Plus, Pro, Team, and Free users, with Enterprise and Edu access following shortly.