Say Hello to GPT-4o Mini: The Budget-Friendly AI Model

Starting today, GPT-4o mini takes the spotlight, ready to revolutionize the AI landscape with its cost-efficiency and stellar performance.

Enterprise users will get their hands on this powerhouse next week, aligning with our mission to democratize AI for everyone.

Let’s dive into what makes GPT-4o mini a game-changer.

Cutting Costs, Not Corners

Let’s talk numbers.

GPT-4o mini is priced at a jaw-dropping 15 cents per million input tokens and 60 cents per million output tokens.

For comparison, GPT-4 had us digging deep into our wallets at $30 per million input tokens and $60 per million output tokens​.

This drastic cost reduction makes GPT-4o mini a perfect fit for businesses and developers looking to leverage advanced AI without breaking the bank.

Small Size, Big Punch

Don’t let the “mini” fool you.

GPT-4o mini significantly outperforms GPT-3.5 Turbo and other small models in various academic benchmarks, including textual intelligence and multimodal reasoning.

It matches GPT-4o in language support and excels in function calling, enabling developers to create applications that interact seamlessly with external systems.

Moreover, it shows improved long-context performance compared to GPT-3.5 Turbo, making it a versatile tool for a wide range of applications.

In key benchmarks, GPT-4o mini demonstrates superior performance. For reasoning tasks, it scores 82.0% on MMLU, surpassing Gemini Flash and Claude Haiku.

It also leads in math and coding proficiency, scoring 87.0% on MGSM and 87.2% on HumanEval, outperforming both Gemini Flash and Claude Haiku. Additionally, in multimodal reasoning, GPT-4o mini achieves a score of 59.4% on MMMU, again outperforming its competitors.

These results highlight GPT-4o mini’s advanced capabilities and its potential to drive innovation in AI applications.

Versatility and Safety

Safety is a cornerstone of GPT-4o mini, built into the model from the initial stages and reinforced throughout its development.

During pre-training, OpenAI filters out undesirable content such as hate speech, adult material, personal data aggregation sites, and spam.

Post-training, the model’s behavior is aligned with OpenAI’s policies using reinforcement learning with human feedback (RLHF) to enhance accuracy and reliability.

The same safety measures applied to GPT-4o are incorporated into GPT-4o mini, evaluated rigorously using automated and human assessments according to OpenAI’s Preparedness Framework, and tested by over 70 external experts in various fields to identify and mitigate potential risks.

Building on these safety foundations, GPT-4o mini incorporates new techniques to further enhance its safety.

One notable method is the instruction hierarchy, which improves the model’s resistance to jailbreaks, prompt injections, and system prompt extractions, ensuring more reliable and safer responses for large-scale applications.

OpenAI remains committed to monitoring the usage of GPT-4o mini, continuously improving its safety measures to address emerging risks, and plans to share detailed findings in the upcoming GPT-4o system card and Preparedness scorecard.

Bringing AI to the Masses

Open AI is not just talking about making AI affordable; it is talking about making it accessible.

GPT-4o mini is available through multiple APIs, including the Assistants API, Chat Completions API, and Batch API.

This ensures that whether you’re a solo developer or a large enterprise, you can integrate this powerful model into your projects seamlessly.

A Peek into the Future

Remember when AI was just a fancy concept in sci-fi movies?

Those days are long gone.

We’ve seen remarkable advancements in AI over the past few years, paired with substantial cost reductions.

For instance, the cost per token of GPT-4o mini has dropped by a whopping 99% since text-davinci-003, a less capable model from 2022.

Open AI’s vision? A future where AI is seamlessly integrated into every app and website.

GPT-4o mini is paving the way for developers to build and scale powerful AI applications more efficiently and affordably.

Open AI is on a mission to make AI more accessible, reliable, and embedded in our daily digital lives, and we’re excited to lead the charge.

What’s Next?

As Open AI continues to innovate and drive down costs, it is committed to enhancing model capabilities.

The future of AI is becoming more accessible and integrated into our daily experiences.

With GPT-4o mini, the future is now, and it’s looking brighter (and cheaper) than ever.

If you want more updates related to AI, subscribe to our Newsletter