OpenAI has officially launched GPT-5, its most advanced and ambitious artificial intelligence model to date. Unveiled on Thursday, GPT-5 marks a significant evolution in AI, blending the deep reasoning capabilities of OpenAI’s o-series with the rapid responsiveness of its GPT-series. This new flagship model powers the next generation of ChatGPT and represents a key milestone in OpenAI’s pursuit of artificial general intelligence (AGI).
“Having something like GPT-5 would be pretty much unimaginable at any previous time in history,” said OpenAI CEO Sam Altman during a press briefing. He hailed the model as “the best in the world” and a “significant step” toward building AI that can outperform humans in economically valuable tasks.
More Than a Chatbot: GPT-5 Acts Like an Agent
While GPT-4 made headlines for its ability to answer complex questions with remarkable fluency, GPT-5 goes further — it can complete tasks on behalf of users. From generating full software applications and scheduling events to drafting in-depth research briefs, GPT-5 positions ChatGPT as a proactive assistant rather than just a conversational tool.
The model features a new real-time router that dynamically chooses the best response mode — balancing speed and depth — removing the need for users to manually adjust settings. This intelligent routing enhances usability for both novice and expert users.
Free and Widely Available
In a bold move, OpenAI is making GPT-5 the default model for all free ChatGPT users, bringing advanced reasoning capabilities to a broader audience for the first time.
“This is just one of the ways that I’m excited to live the mission — making sure that this stuff actually benefits people,” said Nick Turley, VP of ChatGPT, emphasizing OpenAI’s commitment to accessibility.
Performance Benchmarks: Strengths and Trade-Offs
GPT-5 delivers state-of-the-art results in key AI domains, edging out top contenders from Anthropic, Google DeepMind, and xAI in several benchmarks.
Coding and “Vibe Coding”
- On the SWE-bench Verified benchmark (real-world GitHub coding tasks), GPT-5 scored 74.9%, outperforming Anthropic’s Claude Opus 4.1 (74.5%) and DeepMind’s Gemini 2.5 Pro (59.6%).
- Sam Altman highlighted GPT-5’s strength in generating entire software applications — a capability being dubbed “vibe coding.”
Academic Reasoning
- On Humanity’s Last Exam (covering math, humanities, sciences), GPT-5 Pro scored 42% with tools, slightly behind xAI’s Grok 4 Heavy (44.4%).
- On GPQA Diamond, a PhD-level science benchmark, GPT-5 Pro scored 89.4%, beating Claude Opus 4.1 (80.9%) and Grok 4 Heavy (88.9%).
Health Advice Accuracy
- On HealthBench Hard Hallucinations, GPT-5 “with thinking” hallucinated only 1.6% of the time — a drastic improvement over GPT-4o (12.9%) and o3 (15.8%).
Despite some underperformance in specific agentic tasks (e.g., GPT-5 scored 63.5% on navigating an airline website vs. o3’s 64.8%), it demonstrated strong performance in retail site navigation (81.1%).
Reduced Hallucinations, Increased Safety
GPT-5 offers significant improvements in truthfulness and reliability:
- It hallucinates only 4.8% of the time in ChatGPT responses — a major drop from o3 (22%) and GPT-4o (20.6%).
- OpenAI’s safety research lead, Alex Beutel, noted that GPT-5 is less deceptive, more transparent, and better at differentiating malicious queries from harmless ones.
Creative Edge and “Better Vibes”
OpenAI says GPT-5 also excels in subjective and creative domains, including design and writing. Turley remarked that the model shows more natural expression and exhibits what he called “better taste.”
“The vibes of this model are really good,” he said.
New Personalities and Subscription Tiers
As part of the GPT-5 rollout, ChatGPT is getting four new personality modes: Cynic, Robot, Listener, and Nerd, allowing users to tailor interactions to their preferences without additional prompting.
GPT-5 is available across ChatGPT’s subscription tiers:
- Free users: Access to GPT-5 with standard limits
- Plus subscribers ($20/month): Higher usage caps
- Pro subscribers ($200/month): Unlimited access, including GPT-5 Pro, a version with enhanced reasoning via increased computational resources
- Enterprise, Team, and Edu users: Default GPT-5 rollout next week
Developer Access and Pricing
For developers, GPT-5 is available via OpenAI’s API in three variants:
- gpt-5
- gpt-5-mini
- gpt-5-nano
These differ in processing time and depth of reasoning. A new verbosity control feature lets developers set response length based on task needs.
Pricing:
- $1.25 per million input tokens
- $10 per million output tokens
