OpenAI has finally unveiled GPT-5, their latest and most powerful AI model, during a highly anticipated livestream. This launch brings a host of improvements designed to make AI smarter, faster, and more practical for everyday use. Whether you’re a developer, business user, or just curious, here’s everything announced to help you catch up.
The event kicked off with Sam Altman and a team of OpenAI experts demonstrating GPT-5’s capabilities through live demos and benchmarks. They highlighted how the model outperforms predecessors like GPT-4o and even rivals such as Grok in key areas. With nearly 700 million weekly ChatGPT users already, this upgrade aims to integrate deeper intelligence into work and creativity.
One standout theme was GPT-5’s focus on agentic tasks, where the AI acts more autonomously to solve complex problems. Demos showed it handling coding, reasoning, and voice interactions with impressive naturalness. OpenAI emphasised safety and reliability, ensuring the model refuses harmful requests while providing partial answers where appropriate.
Benchmarks revealed GPT-5’s dominance in long-context understanding, maintaining near-perfect accuracy up to 256k tokens, while growing the maximum token window to 400k. It also excelled in tool use for sectors like telecom, retail, and airlines, scoring higher than previous models. Instruction following saw gains too, with better results in freeform writing and multi-turn challenges.
The model family includes GPT-5, GPT-5 mini, and GPT-5 nano, catering to different needs from full power to efficiency. Developers get enhanced API tools, including verbosity controls and parallel tool calling. For general users, the interface sports a fresh purple-themed design, symbolising the model’s vibrant personality.
Key Features of GPT-5
Expanded Context Window
GPT-5 boosts the context window to 400,000 tokens, allowing it to handle massive inputs without losing details. This means better performance on long documents or conversations, as shown in demos where it recalled intricate information accurately.
Advanced Reasoning and Agentic Capabilities
The model shines as the best agentic AI yet, thinking step-by-step and chaining tools for tasks like coding or problem-solving. It scored 100 on the AIME2025 benchmark for Pro users, outperforming others in maths, physics, and health applications.
Improved Coding Assistance
GPT-5 can generate hundreds of lines of code in minutes, complete with an expandable pane showing its thought process. Demos featured it building interactive experiences incorporating maths and physics, making it a top choice for developers.
Enhanced Voice and Study Modes
Voice interactions are more natural, with adjustable speed and a new ‘Study and Learn Mode’ for educational use. Personalities add flair, letting users tailor the AI’s tone for better engagement.
Better Writing and Content Generation
Outputs feel more human-like, reducing AI-generated stiffness for natural, creative writing. Comparative demos against GPT-4o showed clearer, more nuanced responses.
Safety and Refusal Mechanisms
GPT-5 handles edge cases smarter, offering partial answers or explanations instead of outright refusals. It resists promptjacking and explains boundaries, with built-in safeguards against harmful content.
Multimodal and Tool Integration
Supports web search, file handling, and image generation natively. For enterprises, it integrates with tools like Microsoft 365 Copilot, promising faster workflows.
During the event, Sam Altman highlighted its potential in specialised fields.
“GPT-5 is the best model ever for health.”
Developers praised its coding prowess in the announcements.
“GPT-5 is the smartest coding model we’ve used. Our team has found GPT-5 to be remarkably intelligent, easy to steer, and even to have a personality we haven’t seen in any other model.”
Michael Truell, Co-Founder & CEO, Cursor.
Availability starts today for all ChatGPT users on a free tier with usage quotas, exceed them, and it reverts to GPT-5 mini. For unlimited access, upgrade to ChatGPT Pro or higher plans. Team plans offer unlimited messages at A$38 per user per month billed annually or A$45.60 monthly, with generous GPT-5 thinking access.
Enterprises and education users get access next week, including GPT-5 Pro for deeper reasoning. API developers can integrate today, with token pricing at US$1.25 per million input for the full model—check OpenAI’s site for Australian equivalents.
If you missed the livestream, re-watch it on YouTube for the full demos and insights.
For more information, head to https://openai.com/chatgpt