GitHub accidentally leaked details about a new line of OpenAI GPT-5 models in a blog post that was quickly deleted. The post said that GPT-5 would be available in four flavors and offer “significant improvements in logical reasoning, code quality, and user experience,” The Verge reports.
According to an archived version of the post, GPT-5 will receive enhanced agent capabilities and will be able to solve complex programming tasks with minimal prompts. It is known that the line will include the following models:
- gpt-5 is a basic version focused on logic and multi-step tasks.
- gpt-5-mini is a lightweight version for scenarios with a limited budget.
- gpt-5-nano is a super-fast model with minimal latency, designed for local use.
- gpt-5-chat is specially designed for advanced multimodal dialogues and applications in a corporate environment.
The launch of GPT-5 was previously reported in the media, and last month the first rumors appeared that OpenAI was preparing a release in August, including mini and nano versions that would be available via the API.
Earlier this week, OpenAI employees began hinting at a major event, with CEO Sam Altman even joking about GPT-5 on social media. Now, OpenAI has actually confirmed the launch of GPT-5. The LIVE5TREAM event is scheduled for August 7, 2025 at 10:00 AM Pacific Time (20:00 Kyiv time).
Recall that OpenAI recently introduced two open-source models called gpt-oos.