Goodbye GPT-3.5, OpenAI’s New GPT-4o Mini AI Model is All About Compact Power
Introduction of GPT-4o Mini
OpenAI has introduced a new large language model (LLM) called GPT-4o Mini to ChatGPT and its APIs. As the name implies, the GPT-4o Mini model is a more compact version of the GPT-4o model introduced in May. This mini model is designed to balance the power of GPT-4o with a more cost-efficient approach.
Key Features of GPT-4o Mini
GPT-4o Mini retains much of the functionality of its larger counterpart, although the API currently supports only text and vision, with image, video, and audio inputs and outputs still in development. Like GPT-4o, the new model has a context window of 128,000 tokens, which is eight times that of GPT-3.5 Turbo. Additionally, the new model includes enhanced safety features, incorporating new techniques to make it more resistant to jailbreaks and improper prompt injections, addressing developers’ concerns for broad AI API deployment.
Ready for Bigger Jobs
OpenAI suggests that the larger context window and other improvements, such as enhanced non-English text understanding, will make GPT-4o Mini particularly useful for processing large documents or linking 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 exams more quickly than other models. OpenAI envisions GPT-4o Mini automating and streamlining business processes due to its ability to fetch data and take actions with external systems. For businesses using the API, the cost is notably reduced to just over half the price per token of GPT-3.5 Turbo.
Broader Impact and Comparison
OpenAI is committed to making intelligence as broadly accessible as possible. The introduction of GPT-4o Mini is expected to significantly expand the range of applications built with AI by making intelligence much more affordable. GPT-4o Mini is part of a recent wave of smaller LLMs like Google’s Gemini Flash and Anthropic’s Claude Haiku. However, GPT-4o Mini outperforms these models in 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. It also excelled in the MGSM and Human Eval tests, achieving 87% and 87.2%, compared to Gemini Flash’s 75.5% and 71.5%, and Haiku’s 71.7% and 75.9%. GPT-4o Mini demonstrates superior performance in textual comprehension as well as math and coding tasks.
Mini Model Major Plans
The introduction of GPT-4o Mini represents a significant step in making advanced AI more affordable and accessible. Lower costs combined with better performance are likely to facilitate the incorporation of AI into everyday applications. ChatGPT users can access the model starting this week, and OpenAI plans to introduce fine-tuning capabilities for GPT-4o Mini within the API.
Evolution of ChatGPT and Future Plans
This development marks another step in the evolution of ChatGPT’s services. As OpenAI phases out GPT-3.5 for ChatGPT, the focus shifts to providing more powerful models. OpenAI CEO Sam Altman has hinted that GPT-5 will “substantially improve” upon existing models. Meanwhile, the leaked OpenAI scale for measuring AI power indicates that there is still a long way to go to achieve artificial general intelligence (AGI) that can perfectly mimic the workings of the human mind.
