The evolution of artificial intelligence models is proceeding at a rapid pace, as demonstrated by the recent announcement of GPT-4o mini by OpenAI. This new model will replace its predecessor, GPT-3.5, bringing with it a series of significant innovations and improvements. The name is familiar, considering it's a younger (or older?) sibling of the the one presented a short time ago.
Official GPT-4o mini: very powerful but low-cost model from OpenAI
OpenAI described GPT-4o mini as the more efficient and convenient model currently available in its segment. The company also previewed its future plans, which include the integration of advanced features for multimedia content processing, including images, video e audio. This move aligns with the increasingly marked trend towards artificial intelligence systems capable of interpreting and generating content in multiple formats.
The new model will be accessible to all users. Its implementation, however, will take place gradually: initially it will be available to users of the free version of ChatGPT, ChatGPT Plus subscribers and team members, while ChatGPT Enterprise users will be able to benefit from it starting from the following week. One of the most notable features of GPT-4o mini is its ability to process up to 128.000 tokens per context, combined with an updated knowledge base until October 2023.
From an economic point of view, the introduction of GPT-4o mini represents a strategic move for OpenAI. The new model offers a significant advantage in terms of operating costs, with rates of 15 cents per million token input and 60 cents per million token output. These prices translate into significant savings compared to previous models, resulting in the 60% less expensive than GPT-3.5 Turbo.
The performance of GPT-4o mini has been subjected to rigorous comparative tests. According to LMSYS ranking, the new model has surpassed GPT-4 in chat options, demonstrating its effectiveness in conversational contexts. Additionally, GPT-4o mini has proven to excel in academic settings, outperforming GPT-3.5 Turbo in both textual analysis and multimodal reasoning. Another strong point of the new model is its versatility linguistics, which matches that of GPT-4o, allowing interactions in a wide range of languages.
