OpenAI launches GPT-4o Mini: A cheaper, lighter AI model for developers

OpenAI is levelling the playing field for AI development with the launch of GPT-4o Mini, a smaller and significantly more affordable version of its powerful language model. This new offering aims to empower developers who were previously priced out of using larger models like GPT-4.

The high cost of building AI applications with OpenAI's full-sized models has been a significant barrier for many developers. This has led some to turn to cheaper alternatives like Google's Gemini 1.5 Flash and Anthropic's Claude 3 Haiku. Recognising this need for greater accessibility, OpenAI is now offering a more budget-friendly option.

GPT-4o Mini, available to ChatGPT users on Free, Plus, and Team plans starting today, boasts greater capabilities than GPT-3.5 while remaining significantly cheaper to operate. It will also be available to developers via OpenAI's API, offering support for text and vision processing, with plans to expand to multimodal inputs and outputs like video and audio in the near future.

While the GPT-4o Mini is designed for simpler tasks, its performance is impressive. It achieved an 82% score on the MMLU benchmark, surpassing GPT-3.5's 70% score and rivalling the scores of competing light models like Claude 3 Haiku and Gemini 1.5 Flash.

OpenAI showcased the potential of GPT-4o Mini through partnerships with the fintech startup Ramp and the email client Superhuman. Ramp used the model to build a receipt-scanning tool that extracts expense data from images, while Superhuman leveraged it to create an AI-powered email response suggestion feature.