Everything We Know About GPT-5, OpenAI’s Latest Model
OpenAI’s new model, GPT-5, is just around the corner. The company plans to unveil it today in a “longer than usual” presentation , which suggests the company has some exciting news up its sleeve. That’s because it’s already been revealed — at least in part — as some news about GPT-5 has leaked ahead of the announcement.
A now-deleted blog post ( archived here ) shows some of the new features and changes included in GPT-5. Unsurprisingly, the post calls GPT-5 “OpenAI’s most advanced model,” and unsurprisingly, it improves “intelligence, code quality, and user experience” over previous models. However, the post does say that the model can create and execute code with “minimal guidance,” and that it now has “improved agent capabilities.”
Companies are pushing agent AI more than ever: Instead of simply generating results like text, images, or videos based on queries, agent AI can perform tasks on behalf of users. For example, ChatGPT’s agent mode can take control of your web browser to perform tasks, or create an Excel document based on a query for a specific spreadsheet.
The publication does not provide details on the performance and functionality improvements of GPT-5 compared to previous models. At this point, we do not have any diagrams or images to work with. However, the publication does say that GPT-5 is divided into four separate models depending on the use case. These include:
-
GPT-5 : This model is designed for logical and multi-step tasks.
-
GPT-5-mini : This model is not as powerful as GPT-5 and as such is intended for “cost-sensitive” applications.
-
GPT-5-nano : This model is designed to run fast in “low latency” applications.
-
GPT-5-chat : This is an OpenAI model that allows companies to perform multimodal and context-aware transformations.
In addition to this GitHub post, BleepingComputer reports that it has found documents suggesting an OpenAI pricing plan for GPT-5. As with previous models, users will be able to access GPT-5 for free. However, ChatGPT Plus subscribers can access GPT-5 with “advanced reasoning features,” while ChatGPT Pro users can get “professional reasoning features.” Reasoning models take more time to “think” about a query, breaking the query down into individual steps before processing it. BleepingComputer claims that GPT-5 can automatically sync between reasoning features and no reasoning features, but you can nudge it to perform these features at any time.
That’s all we know about GPT-5 at the time of writing. But we won’t have long to wait: the leak came just hours before OpenAI officially announced its newest model, so it won’t be long before we compare GPT-5 to other models like GPT-4o , Google’s Gemini , and Anthropic’s Claude .
How to watch the OpenAI GPT-5 presentation
You can watch the announcement live today at 10:00 AM PT (1:00 PM ET). To watch, head to the official OpenAI YouTube channel and click on the GPT-5 live stream.
Disclosure: Lifehacker’s parent company, Ziff Davis, filed a lawsuit against OpenAI in April, alleging that it infringed Ziff Davis’ copyrights in the training and operation of its AI systems.