Meet GPT-5, the latest leap in AI intelligence that brings expert-level capabilities right to your fingertips. This unified system is designed to be smarter, faster, and more useful than any previous model, with built-in thinking that adapts to your needs. Whether you’re coding, writing, or seeking health insights, GPT-5 delivers state-of-the-art performance across the board.
At its core, GPT-5 operates as a unified system with three key components: a smart, efficient model for quick answers, a deeper reasoning model (GPT-5 thinking) for complex problems, and a real-time router that intelligently selects the best approach based on conversation type, complexity, and your explicit instructions. This router continuously learns from user interactions, improving over time to provide more accurate and helpful responses.
When it comes to coding, GPT-5 stands out as the strongest model yet. It excels in complex front-end generation and debugging larger repositories, often creating beautiful, responsive websites, apps, and games from a single prompt. Early testers have praised its intuitive design choices, including better understanding of spacing, typography, and white space.
For creative expression and writing, GPT-5 serves as your most capable collaborator. It helps transform rough ideas into compelling, resonant writing with literary depth and rhythm, reliably handling structural ambiguity like unrhymed iambic pentameter or free verse. These improvements make everyday tasks like drafting reports, emails, and memos more efficient and polished.
In health-related queries, GPT-5 empowers users to be informed advocates for their well-being. It scores significantly higher on HealthBench, acting as an active thought partner that proactively flags concerns and asks clarifying questions. While it doesn’t replace medical professionals, it helps you understand results, ask the right questions, and weigh options during decision-making.
Evaluations show GPT-5 is much smarter across academic and real-world benchmarks, achieving state-of-the-art results in math (94.6% on AIME 2025 without tools), coding (74.9% on SWE-bench Verified), multimodal understanding (84.2% on MMMU), and health (46.2% on HealthBench Hard). With GPT-5 Pro’s extended reasoning, it sets new records on GPQA, scoring 88.4% without tools.
The model demonstrates significant gains in instruction following and agentic tool use, reliably carrying out multi-step requests and coordinating across different tools. Its multimodal capabilities allow accurate reasoning over images, charts, and diagrams, while its performance on economically valuable tasks matches or exceeds experts in roughly half the cases across over 40 occupations.
GPT-5 delivers faster, more efficient thinking, performing better than previous models with 50-80% fewer output tokens across visual reasoning, agentic coding, and graduate-level scientific problem solving. It’s also significantly less likely to hallucinate, with responses ~45% less likely to contain factual errors than GPT-4o, and ~80% less likely when thinking.
Safety takes a major step forward with GPT-5’s new “safe completions” approach, providing more nuanced responses while staying within safety boundaries. The model is less sycophantic, uses fewer unnecessary emojis, and feels more like chatting with a helpful, PhD-level friend than talking to AI.
Customization options expand with four new preset personalities—Cynic, Robot, Listener, and Nerd—letting you tailor ChatGPT’s interaction style without writing custom prompts. All meet strict internal evaluations for reducing sycophancy while maintaining helpfulness.
For the most challenging tasks, GPT-5 Pro offers extended reasoning capabilities, making 22% fewer major errors and excelling in health, science, mathematics, and coding. External experts preferred GPT-5 Pro over “GPT-5 thinking” 67.8% of the time in evaluations.
GPT-5 is now the default model in ChatGPT, automatically applying reasoning when beneficial. Paid users can select “GPT-5 Thinking” from the model picker or use prompts like “think hard about this” to ensure reasoning is used. The rollout begins today for Plus, Pro, Team, and Free users, with Enterprise and Edu access coming next week. Usage limits vary by subscription level, with free users transitioning to GPT-5 mini after reaching their limits.
