OpenAI has unveiled its latest flagship model, GPT-5, designed to power the next generation of ChatGPT and significantly expand its capabilities.
Released on Thursday (7 August 2025), GPT-5 is the company’s first “unified” model, meaning it combines the reasoning power of the “o-series” with the speed and responsiveness of the GPT series. This fusion allows ChatGPT to act more like an AI agent rather than just a simple chatbot.
Unlike GPT-4, which primarily focused on answering questions, GPT-5 can handle a wider variety of tasks, including:
- Writing and developing software
- Managing a user’s calendar
- Producing research summaries
Savy (Chief Editor) Aug 13, 2023 • 3 min read
Real-Time Router: Smarter Response Decisions
A key upgrade is the real-time router — a smart decision-making system that determines whether GPT-5 should respond instantly or take extra time to “think” for a more thorough answer. Users no longer need to choose the response style themselves; the model decides automatically.
CEO’s Statement
OpenAI CEO Sam Altman described GPT-5 as:
“The best model in the world and a significant step toward Artificial General Intelligence (AGI).”
Altman noted that having a model like GPT-5 in human history was “almost unimaginable.”
Free Access for All ChatGPT Users
Starting Thursday, GPT-5 is available by default for all free ChatGPT users — a major shift from earlier policies where advanced models were reserved for paying subscribers.
ChatGPT VP Nick Turley said the goal is to make advanced AI models accessible to everyone.
Performance & Competitor Comparison
According to OpenAI, GPT-5 performs at the level of the world’s most advanced AI systems — and in some areas, it outperforms them:
Notably, GPT5 also delivers more natural and creative writing, design, and idea generation.
| Metric | GPT-5 Score | Closest Competitor |
|---|---|---|
| SWE-bench Verified (real-world coding) | 74.9% | Claude Opus 4.1 – 74.5% |
| GPQA Diamond (PhD-level science questions) | 89.4% | Grok 4 Heavy – 88.9% |
| HealthBench Hard Hallucinations (medical accuracy) | 1.6% error | GPT-4o – 12.9% error |
Greater Accuracy, Stronger Safety, and More Personalization
GPT-5 is not only faster and smarter — it’s more accurate than ever. In OpenAI tests, the model produced incorrect or “hallucinated” information only 4.8% of the time, compared to around 20% for GPT-4o. That means users can trust its outputs more consistently.
It’s also safer and more transparent, showing less inclination toward dangerous or misleading responses. GPT5 is better at distinguishing between normal users and those attempting malicious use.
OpenAI has also added four new ChatGPT personalities to make conversations more fun and tailored:
- Cynic – skeptical and dry-witted
- Robot – formal and mechanical
- Listener – empathetic and attentive
- Nerd – deeply knowledgeable and enthusiastic
These personalities change ChatGPT’s tone and style automatically — no extra prompts required. Free users get access to GPT-5, while Plus and Pro subscribers enjoy fewer limits and the upgraded GPT-5 Pro version.
API for Developers
GPT-5 is available in three sizes: gpt-5, gpt-5-mini, and gpt-5-nano.
Pricing of GPT-5:
- Input: $1.25 per million tokens
- Output: $10 per million tokens
Bottom line: GPT5 marks a major step forward in AI. While its lead over competitors isn’t massive in every category, its advancements in reasoning, accuracy, and safety could transform the way people use ChatGPT.
Here you can see our other blogs about new AIs:
Related blogs
