Goodbye GPT-3.5, OpenAI’s new GPT-4o mini AI model is all about compact power
OpenAI has added a new large language model (LLM) called GPT-4o mini to ChatGPT and its APIs. As the name suggests, the GPT-4o Mini model is a smaller version of the GPT-4o model introduced in May. The mini model is designed to balance the power of GPT-4o with a more cost-efficient approach.
GPT-4o mini retains much of the functionality of its larger cousin, although its API only supports text and image for now, with image, video, and audio input and output still in the works. Like GPT-4o, the new model has a context window of 128,000 tokens, or eight times as much as GPT-3.5 Turbo. The new model also comes with improved security features. In addition to the features already built into GPT-4o, GPT-4o mini adds new techniques that make it more resistant to jailbreaks and false prompt injections, among other concerns that developers looking to widely implement AI APIs.
Ready for bigger jobs
OpenAI suggests that the larger context window and other upgrades, such as improved understanding of non-English text, will make GPT-4o mini particularly useful for processing large documents or pairing multiple interactions with the AI model. For example, it could provide better recommendations in online stores, speed up real-time text responses for customer service, and produce accurate and detailed answers for students studying for an exam faster than other models. OpenAI has visions of GPT-4o automating and streamlining business processes through its ability to retrieve data and take actions with external systems. For businesses that use the API, the cost is significantly reduced to just over half the per-token price of GPT-3.5 Turbo.
“OpenAI aims to make intelligence as widely accessible as possible,” OpenAI said. explained in his announcement. “We expect that GPT-4o mini will significantly expand the range of applications built with AI by making intelligence much more affordable.”
GPT-4o mini is part of the recent wave of smaller LLMs like Google’s Gemini Flash and Anthropic’s Claude Haiku. However, according to OpenAI, GPT-4o mini blows them out of the water when it comes to many of the standard tests. The model scored 82% on the Massive Multitask Language Understanding (MMLU) benchmark, compared to 77.9% and 73.8% by Gemini Flash and Haiku, respectively. The same goes for the MGSM and Human Eval tests, where GPT-4o Mini managed 87% and 87.2%, while Gemini Flash had 75.5% and 71.5%, and Haiku had 71.7% and 75.9%. In other words, GPT-4o Mini wins on textual comprehension in addition to math and coding tasks, as can be seen in the chart below.
Mini Model Master Plans
The introduction of GPT-4o Mini is an important step in making advanced AI more affordable and accessible, OpenAI says. Lower costs plus improved performance will likely help integrate AI into everyday applications. The same goes for ChatGPT users, who all have access to the model starting this week. OpenAI also plans to introduce fine-tuning capabilities for GPT-4o Mini within the API.
The broader picture shows another step in ChatGPT’s evolving services. As OpenAI phases out GPT-3.5 for ChatGPT, the focus shifts to the next phase of delivering more powerful models. OpenAI CEO Sam Altman has long hinted at how GPT-5 will “substantially improve” existing models. At the same time, the leaked OpenAI scale for measuring AI power shows there’s still a long way to go toward the still-mythical artificial general intelligence (AGI) that can perfectly mimic the workings of the human mind.
