On August 7, 2025, OpenAI officially launched GPT-5, the newest version of its renowned language model. This model is accessible to all ChatGPT users, both free and subscription-based. CEO Sam Altman described GPT-5 as “a significant step along the path to AGI,” noting that while it does not yet embody true artificial general intelligence, it demonstrates a higher level of intelligence than its predecessors.
Key Features of GPT-5
OpenAI claims that GPT-5 is:
- Smarter: Enhanced cognitive abilities that provide deeper insights.
- Faster: An improved processing speed for quicker responses.
- More Useful: Additional functionalities that enhance user experience.
- More Accurate: Lower rates of hallucination when generating responses.
Altman highlighted that conversing with GPT-5 feels similar to engaging with an expert, comparable to a PhD-level educated individual in various subjects.
New Models Introduced
With the launch, OpenAI introduced two variants:
- GPT-5 Mini: A lightweight model for users requiring less input.
- GPT-5 Nano: An API-only version that provides cost-effective options for developers.
Free users will have access to both GPT-5 and GPT-5 Mini, while the Plus subscription offers extended usage limits. The Pro tier, priced at $200 monthly, provides unlimited access to GPT-5, GPT-5 Pro, and a thinking model that processes queries more efficiently.
API Pricing
The new pricing structure for API users is as follows:
- $1.25 for every 1 million input tokens and $10 for output tokens (GPT-5).
- $0.25 for every 1 million input tokens and $2 for output tokens (GPT-5 Mini).
- $0.05 for every 1 million input tokens and $0.40 for output tokens (GPT-5 Nano).
Developers will find GPT-5 Nano particularly attractive due to its affordability compared to popular alternatives like Gemini 2.5 Flash.
Integration and User Experience
In the coming week, Pro users can integrate their Gmail, Google Contacts, and Google Calendar with ChatGPT for a more seamless experience. Upcoming features will include:
- Automatic referencing of integrated data during chats.
- Customizable chat colors.
- Personality presets including: Cynic, Robot, Listener, and Nerd.
These features are designed to elevate user interaction, making it richer and more relatable.
Enhanced Contextual Understanding
GPT-5 boasts a 256,000-token context window, allowing it to handle extended conversations, long documents, or complex codes without losing track of the topic. The performance on coding benchmarks has also shown significant improvement, achieving higher scores across various coding tasks:
- SWE-Bench Verified: 74.9%
- SWE-Lancer: 55%
- Aider Polyglot: 88%
These benchmarks underscore GPT-5’s coding capabilities, showcasing its adeptness at bug fixing and multi-programming languages.
Health-Related Queries
OpenAI emphasizes that GPT-5 is a top performer for health-related inquiries. It scored significantly higher than previous models in health benchmarks:
- HealthBench Hard: Scored 25.5% vs. o3’s 31.6%.
These improvements are vetted by health professionals, assuring users of its reliability in sensitive matters.
Safety and Reliability Enhancements
OpenAI has incorporated strategies to lessen factual inaccuracies. The hallucination rate was reduced by 26% compared to GPT-4, with the thinking variant reflecting a 65% decrease in errors. Their safety protocols underwent rigorous testing, including over 5,000 hours of red teaming.
Growing User Base
OpenAI continues to expand its reach, currently boasting nearly 700 million weekly active users for ChatGPT, including 5 million paying business customers and 4 million developers utilizing the API. The head of ChatGPT, Nick Turley, commented on the positive feedback of this model, anticipating a favorable reception among general users.
With the launch of GPT-5, OpenAI has marked a notable advancement in AI capabilities. Users can expect a more intelligent, faster, and accurate model that significantly enhances user experience and reliability.
