OpenAI Launches GPT-4o Mini: A More Affordable, Compact AI Model

OpenAI unveiled its latest AI model, GPT-4o mini, on Thursday. This compact model is designed to be faster and more cost-effective than OpenAI’s existing top-tier AI solutions. Starting today, developers can access GPT-4o mini, and it will also be available via the ChatGPT web and mobile apps for consumers, with enterprise access rolling out next week.

According to OpenAI, GPT-4o mini excels in reasoning tasks that involve both text and vision, outperforming other leading small AI models. As smaller AI models gain traction among developers, their speed and cost advantages make them an attractive choice for high-volume, repetitive tasks, distinguishing them from larger models like GPT-4 Omni and Claude 3.5 Sonnet.

Replacing GPT-3.5 Turbo as OpenAI’s smallest model, GPT-4o mini boasts an impressive MMLU benchmark score of 82%, surpassing Gemini 1.5 Flash at 79% and Claude 3 Haiku at 75%, as reported by Artificial Analysis. Additionally, it achieved a score of 87% on MGSM, a math reasoning benchmark, compared to Flash's 78% and Haiku's 72%.

OpenAI emphasizes that GPT-4o mini is significantly more economical to operate, being over 60% cheaper than GPT-3.5 Turbo. Currently, the model supports both text and vision capabilities via API, with future plans to add video and audio functionalities.

Olivier Godement, OpenAI’s Head of Product API, stated in an interview, "To empower every corner of the world with AI, we must lower the cost of our models. GPT-4o mini represents a significant advancement in that effort."

For developers utilizing OpenAI’s API, GPT-4o mini is priced at $0.15 per million input tokens and $0.60 per million output tokens. The model features a context window of 128,000 tokens, equivalent to the length of a typical book, and has a knowledge cut-off in October 2023.

While OpenAI hasn’t disclosed the exact size of GPT-4o mini, it is positioned similarly to other small AI models, like Llama 3 8b, Claude Haiku, and Gemini 1.5 Flash. Nevertheless, OpenAI asserts that GPT-4o mini is faster, more cost-efficient, and smarter than its peers, based on pre-launch testing in the LMSYS.org chatbot arena, which early independent tests support.

George Cameron, Co-Founder at Artificial Analysis, noted, “Compared to similar models, GPT-4o mini is remarkably fast, with a median output speed of 202 tokens per second. This speed more than doubles that of GPT-4o and GPT-3.5 Turbo, making it an excellent choice for speed-sensitive applications, including various consumer use cases.”

OpenAI's New Tools for ChatGPT Enterprise

In addition to GPT-4o mini, OpenAI has introduced new features for enterprise clients. In a recent blog post, OpenAI launched the Enterprise Compliance API, designed for businesses operating in heavily regulated sectors like finance, healthcare, legal services, and government. This API aims to assist these organizations in meeting logging and audit requirements.

The new tools empower administrators to track and manage their ChatGPT Enterprise data better. The API offers detailed logs of time-stamped interactions, covering conversations, uploaded files, workspace users, and more.

OpenAI has also enhanced control for workspace GPTs—customized versions of ChatGPT tailored to specific business needs. Previously, admins could only fully allow or block GPT actions but can now curate a list of approved domains for GPT interaction.

By providing innovative solutions like GPT-4o mini and enhanced enterprise tools, OpenAI continues to solidify its position as a leader in the AI landscape, driving performance and affordability for developers and businesses alike.