OpenAI has announced GPT-4o mini, a small yet powerful AI model designed to make artificial intelligence more affordable and accessible. Priced at just 15 cents per million input tokens and 60 cents per million output tokens, GPT-4o mini is significantly cheaper than previous models such as GPT-3.5 Turbo, providing a cost-effective solution for a variety of applications.
According to the company's blog, GPT-4o mini boasts strong performance metrics, including scoring 82% on the MMLU benchmark and outperforming GPT-41 on the LMSYS leaderboard in a chat setting. The model's low cost and fast response time enable it to handle a wide range of tasks, making it ideal for applications that require multiple model invocations, large amounts of context, or real-time text interaction, such as customer support chatbots.
Currently, GPT-4o mini supports text and vision input, and will support image, video, and audio input and output in the future, OpenAI said in a blog post. It has a context window of 128K tokens, supports up to 16K output tokens per request, and knowledge will be updated until October 2023. An improved tokenizer makes processing non-English text more cost-effective.
The model excels in both academic and practical applications, outperforming other mini models in reasoning, mathematics, and coding tasks. For example, GPT-4o mini scored 87% in mathematical reasoning and 87.2% in coding performance on benchmarks such as MGSM and HumanEval.
OpenAI has built robust safety measures into GPT-4o mini, including filtering out harmful content during pre-training and using reinforcement learning with human feedback (RLHF) to align model behavior with safety policies. More than 70 external experts test the model to identify and mitigate potential risks to ensure reliability and safety.
GPT-4o mini is currently available through the Assistant, Chat Completion, and Batch APIs. Access is available to ChatGPT's Free, Plus, and Team users starting today, and to Enterprise users starting next week. OpenAI plans to introduce fine-tuning features for GPT-4o mini soon.
OpenAI aims to continue reducing costs while enhancing AI capabilities, and the introduction of GPT-4o mini is a step towards integrating powerful AI into everyday applications, making advanced intelligence more accessible and affordable for both developers and users.
