OpenAI has announced the release of GPT-4o mini, a more affordable and streamlined version of its flagship AI model, GPT-4o, according to a report by CNBC. The new model aims to cater to a broader range of developers and businesses in the competitive AI services market.
GPT-4o mini will be available to free users and ChatGPT Plus and Team subscribers starting today, with enterprise customers gaining access next week. This model will replace GPT-3.5 Turbo in ChatGPT, providing updated functionality at a lower cost.
GPT-4o’s announcement of the mini AI model is a key part of OpenAI’s initiative to lead in “multimodality,” integrating various forms of AI-generated media—text, images, audio, and video—within the ChatGPT platform.
OpenAI introduced GPT-4o in May, highlighting its ability to process audio and visual information in real time. While some of these features are still pending release due to safety concerns, GPT-4o mini offers comparable capabilities, with plans to expand its functionality over time.
AI companies, including Anthropic and Alphabet’s Google, often release smaller, less advanced versions of their models to provide developers with more options. Smaller models are suitable for high-volume, basic tasks, while larger models handle more complex work. Developers may choose to use both within a single application.
“In our mission to enable the bleeding edge, we want to continue developing frontier models while also offering the best small models,” said Olivier Godement, head of product for OpenAI’s API, in an interview with Bloomberg News. Over the past week, some developers have tested GPT-4o mini, with email startup Superhuman using it for automated replies and financial services startup Ramp using it to extract information from receipts.
Initially, GPT-4o mini will handle text and image inputs and outputs. OpenAI plans to expand its capabilities to process other types of content in the future.GPT-4o mini is also the first model to incorporate a new safety approach called “instruction hierarchy,” which prioritises certain instructions over others. This tactic aims to prevent the AI from performing undesirable actions by giving precedence to directives from organisations.