OpenAI has been teasing something, and it looks like the surprise didn’t quite go as planned. Just before its official event, GitHub accidentally published a blog post revealing details about GPT-5, OpenAI’s next AI model.
The post was quickly deleted, but not before Reddit users spotted it, saved it, and started talking. One user even shared an archived version, making sure the leak lived on even after it was taken down.
The deleted GitHub blog confirmed what many had been speculating: GPT-5 is coming, and it’s not just one model. It’s a whole lineup of four, each tailored for a different use case.
The post also called GPT-5 OpenAI’s “most advanced model” yet, promising major leaps in reasoning, code generation, and how the model interacts with users.
Before it vanished, the blog post listed four variants of GPT-5:
- gpt-5: The main model, designed for logic and multi-step problem solving.
- gpt-5-mini: A lightweight version built for cheaper use and faster responses.
- gpt-5-nano: Even smaller and optimised for low-latency tasks like real-time chat.
- gpt-5-chat: A more conversational model for enterprise apps, with support for natural, multimodal, and highly context-aware dialogue.
The post stated GPT-5 could “handle complex coding tasks with minimal prompting” and described its “enhanced agentic capabilities,” hinting that it may be smarter at taking initiative during tasks.
The leak dropped just before OpenAI’s big reveal. And yes, the event teaser stated “LIVE5TREAM”, with a big bold 5 right in the middle. Moreover, OpenAI CEO Sam Altman hinted that “something big” was coming.
Even before the GitHub post slipped out, some users had already spotted signs in OpenAI’s API docs. Icons for “gpt-5-mini” and “gpt-5-nano” quietly showed up, pointing to an upcoming launch.
