When you purchase through links on our site, we may earn an affiliate commission. This doesn’t affect our editorial independence.
OpenAI has officially launched GPT-5, calling it its most advanced AI model to date. The release signifies a new era for ChatGPT, with the model blending reasoning skills from OpenAI’s o-series with the rapid responses of its GPT series. Furthermore, this combination establishes a more capable and versatile AI assistant.
However, unlike GPT-4 which was already powerful in answering complex questions, GPT-5 can perform more practical tasks. The model can create software applications and manage schedules. In addition, it can also prepare detailed research briefs without heavy user input. This development signals OpenAI’s shift toward building AI agents that can do more than chat.
GPT-5: A Simpler, Smarter User Experience
OpenAI has designed the new model to be easier for everyone to use. The model includes a real-time router that decides how best to respond. Moreover, it can quickly answer straightforward queries or take extra time to produce more thoughtful answers. Users no longer need to adjust settings to get different types of responses.
In Case You Missed It:
During a press briefing, OpenAI CEO Sam Altman called GPT-5 “the best model in the world.” He described the development as a significant milestone toward achieving artificial general intelligence (AGI). According to Altman, having this level of AI capability would have been unimaginable just a few years ago.
From launch day, GPT-5 is the default model for all free ChatGPT users. This is the first time OpenAI has made a reasoning-based AI model freely available. Nick Turley, OpenAI’s VP of ChatGPT, said the move aligns with the company’s mission to put advanced AI in as many hands as possible.
GPT-5’s Rising Expectations and Industry Attention
The release is one of the most anticipated in OpenAI’s history. ChatGPT now attracts over 700 million weekly users, representing nearly 10% of the global population. Many industry watchers see GPT-5 as a key indicator of how far AI technology has progressed.
Technology leaders, investors, and policymakers are watching closely. They want to know if GPT-5 delivers a real leap in capability, similar to GPT-4’s debut in 2023. That launch reshaped expectations of what software could accomplish, especially in professional and creative settings.
Performance Benchmarks and Competitive Standing
OpenAI says the model sets a new performance standard in several key areas. On SWE-bench Verified, a coding benchmark from real-world GitHub tasks, GPT-5 scored 74.9% on its first try. That narrowly beat Anthropic’s Claude Opus 4.1 at 74.5% and far surpassed Google DeepMind’s Gemini 2.5 Pro at 59.6%.
In advanced science knowledge, the Pro version scored 89.4% on GPQA Diamond, a PhD-level exam. It outperformed Claude Opus 4.1’s 80.9% and edged past Grok 4 Heavy’s 88.9%.
However, the version in this discussion does not lead in every area. On Humanity’s Last Exam, which measures performance across multiple academic disciplines, GPT-5 Pro scored 42% with tool use. Grok 4 Heavy achieved a slightly higher 44.4%.
Improved Health Information Accuracy
OpenAI claims GPT-5 is significantly better at providing accurate health-related information. On HealthBench Hard Hallucinations, the model’s hallucination rate was only 1.6%. This is a significant improvement over GPT-4o’s 12.9% and the o3 model’s 15.8%.
Given that millions already turn to AI for health advice, this improvement matters to address previous odd events. Additionally, the new model is more proactive about flagging possible health concerns and clearly explaining medical information.
Stronger Creativity and “Better Taste”
Turley said GPT-5 excels in creative tasks like writing and design, describing it as having “better taste” than rivals. The model’s responses are more natural, and the tone often feels more human.
These creative gains are paired with better factual accuracy. On tests of general response accuracy, GPT-5 with extended thinking mode had an error rate of only 4.8%. By comparison, GPT-4o had 20.6%, and o3 had 22%.
Mixed Results on Agentic Abilities
When tested on the Tau-bench, which simulates online tasks, GPT-5 showed uneven results. In navigating airline websites, it scored 63.5%, slightly below o3’s 64.8%. On retail website navigation, GPT-5 scored 81.1%, just under Claude Opus 4.1’s 82.4%.
Although not dominant in every test, OpenAI asserts that the new model remains a safer and more trustworthy model. Safety research lead Alex Beutel said the AI is less deceptive and better at telling safe from unsafe requests. This reduces unnecessary refusals while still blocking harmful content.
New Features for ChatGPT Users
The GPT-5 rollout includes new personalization options for ChatGPT. Users can choose from four built-in personalities: Cynic, Robot, Listener, and Nerd. These adjust the model’s tone without the user needing to give extra instructions.
Free users now have access to GPT-5, while Plus subscribers get higher usage limits. Pro subscribers, who pay $200 monthly, gain unlimited GPT-5 access along with GPT-5 Pro. This enhanced version uses more computational resources for deeper reasoning. Organizational customers on OpenAI’s Team, Edu, and Enterprise plans will see GPT-5 as their default model starting next week.
Developer Access and Pricing of GPT-5
Developers can now use GPT-5 in the OpenAI API. It comes in three sizes: gpt-5, gpt-5-mini, and gpt-5-nano. These differ in how much time they spend reasoning before responding.
Pricing starts at $1.25 per million input tokens and $10 per million output tokens. This means developers can process large amounts of text at competitive rates. A new API feature also lets developers control verbosity. This allows them to decide whether responses should be concise or detailed, depending on their needs.
Position in the AI Space
The GPT-5 launch follows OpenAI’s release of gpt-oss, an open-weight reasoning model. While gpt-oss nearly matched older models like o3, GPT-5 sets a higher bar in specific domains, especially software development.
The latest update is equal to other top AI models in several areas. Benchmarks alone cannot capture how it will perform in real-world use. Developers and businesses will now test their capabilities in practical scenarios.
Looking Ahead
For OpenAI, the newest version is beyond just another model upgrade. It represents a step toward AI that can manage complex, multi-step tasks for people. Similarly, it reflects the company’s focus on making advanced AI tools widely accessible.
As millions begin using the new model, its real-world impact will come into focus. Success will depend not just on benchmarks but on whether people find it reliable, safe, and genuinely helpful. How the model performs outside of controlled tests may determine its lasting place in the AI space.
