OpenAI released GPT-5 yesterday and to test it out I had it write an article about itslef, based on the research at this link. I then prompted it to make the 'listicle' it wrote more like an article. Here is the article in full.
OpenAI has unveiled GPT-5, calling it the company’s “smartest, fastest, most useful” AI model so far. The release marks a major leap in artificial intelligence capabilities, with improvements across reasoning, factual accuracy, coding, health advice, and creative expression.
Available immediately to all ChatGPT users—with varying access levels depending on subscription—GPT-5 represents a unification of OpenAI’s previous models into a single, adaptable system.
One Model, Multiple Modes
Unlike past versions, GPT-5 operates as a unified system.
It includes:
-
A fast-response model for everyday questions.
-
GPT-5 Thinking, a deeper reasoning mode for complex problems.
-
A real-time router that decides which mode to use based on the difficulty of the task, user instructions, and context.
When a user reaches their usage cap, a GPT-5 Mini model steps in, ensuring uninterrupted service.
Benchmark-Topping Performance
In independent and internal testing, GPT-5 has set new records:
-
Math: 94.6% on the 2025 AIME exam (without tools).
-
Coding: 74.9% on SWE-bench Verified, 88% on Aider Polyglot.
-
Multimodal reasoning: 84.2% on MMMU.
-
Health: 46.2% on HealthBench Hard.
These improvements translate to faster, more accurate results in everyday use—whether that’s writing a wedding toast, diagnosing a complex coding bug, or interpreting a medical lab result.
A Versatile Tool for Work and Creativity
Coding
GPT-5 is the strongest coding model OpenAI has released. It can generate complex, responsive websites and apps in one go, while showing a refined eye for visual design—spacing, typography, and layout. It also handles debugging large repositories with more efficiency than previous models.
Writing
For creative and professional writing, GPT-5 offers richer vocabulary, better structural awareness, and more impactful metaphors. In side-by-side comparisons, it produced poetry with greater emotional depth and vivid imagery than GPT-4o.
Health
In health queries, GPT-5 acts as an informed partner, helping users understand symptoms, results, and treatment options. OpenAI emphasizes that it’s not a substitute for medical professionals, but notes that it’s far more context-aware and precise than any prior model.
Fewer Hallucinations, More Honesty
GPT-5 reduces factual errors dramatically—45% fewer than GPT-4o, and up to 80% fewer when using its reasoning mode. It also detects and communicates limitations more openly, avoiding the overconfident false answers that plagued earlier models.
OpenAI has also tackled sycophancy—the tendency for AI to over-agree with users. GPT-5’s agreeable-but-empty responses have been cut from 14.5% of cases in GPT-4o to under 6%.
A Safer Approach to Sensitive Topics
With GPT-5, OpenAI is introducing “safe completions”, a new safety strategy. Instead of simply refusing to answer sensitive or dual-use questions (such as those in biotechnology), GPT-5 provides partial, high-level, or safe alternative responses when possible, while explaining any refusals clearly.
For biological and chemical domains, OpenAI has implemented multi-layered safeguards—including threat modeling, safe completions, and always-on monitoring—treating GPT-5 Thinking as “high capability” under its Preparedness Framework.
Personalization Gets a Boost
GPT-5 is better at following custom instructions and now offers four built-in personalities: Cynic, Robot, Listener, and Nerd. These presets adjust the AI’s tone and interaction style, making it easier to match different users’ preferences.
GPT-5 Pro for Power Users
For the most complex tasks, OpenAI is releasing GPT-5 Pro, a variant with extended reasoning capabilities. In expert reviews, GPT-5 Pro was preferred over standard GPT-5 Thinking in 68% of cases, making 22% fewer major errors and excelling in science, mathematics, coding, and health.
Availability
GPT-5 is now the default model for all signed-in ChatGPT users, replacing GPT-4o and earlier versions.
-
Free users get limited daily access before switching to GPT-5 Mini.
-
Plus subscribers enjoy higher limits.
-
Pro users receive unlimited GPT-5 and exclusive GPT-5 Pro access.
-
Team, Enterprise, and Education customers get expanded usage limits to support organizational needs.
With GPT-5, OpenAI is positioning ChatGPT as a more reliable, flexible, and creative assistant—one that’s not only smarter, but also more honest and context-aware than any AI it has released before.
Please make the comments constructive, and vulgarity will not be tolerated!