According to OpenAI, this model is faster and more economical than existing AI models that prioritize efficiency.
It is available to both developers and general consumers via the ChatGPT web and mobile applications. Corporate clients will gain access to it later this week.
GPT-4o mini is set to replace GPT-3.5 Turbo as OpenAI’s smallest AI model. OpenAI claims that the new model achieved an 82% score on the MMLU AI benchmark, outperforming Google’s Gemini 1.5 Flash at 79% and Claude 3 Haiku at 75%. On the MGSM benchmark, which evaluates mathematical reasoning, GPT-4o mini scored 87%, compared to 78% for Gemini and 72% for Haiku.
In addition to its superior performance, GPT-4o mini is expected to be significantly more affordable to operate, boasting a cost reduction of over 60% compared to GPT-3.5 Turbo. For developers, the pricing for GPT-4o mini is set at 15 cents per million input tokens and 60 cents per million output tokens. The model also features a context window of 128,000 tokens, equivalent to the length of a book.
Currently, the model supports only text and visual APIs, but its creators have announced plans to include video and audio functionalities in the future. A key highlight of the GPT-4o mini is its enhanced energy efficiency, making it more environmentally friendly than both its predecessors and competitors.
The high energy consumption of modern AI tools has a significant environmental impact. The more energy-efficient GPT-4o mini will now replace the older, more energy-intensive GPT-3.5 Turbo in ChatGPT for users of the free ChatGPT version, as well as the Plus and Team versions.
