OpenAI has recently announced the release of its smaller and more cost-efficient AI model, GPT-4o mini. This model is a direct successor to GPT-3.5 Turbo and offers additional capabilities such as handling text and vision inputs. GPT-4o mini is specifically targeted towards enterprises, startups, and developers building various types of agents, from customer support to financial agents.
The cost of using GPT-4o mini is significantly lower compared to its predecessor. Developers will only have to pay $0.15 USD per 1 million tokens inputted and $0.60 for every million tokens received back from the model. In contrast, GPT-4o costs $5.00 for 1 million input tokens and $15 per 1 million output tokens.
The introduction of GPT-4o mini is seen as a strategic move by OpenAI to maintain its position as a leader in enterprise-grade AI, especially with Meta’s reported release of Llama 3 400-billion parameter model expected soon. By offering a more affordable and capable model, OpenAI aims to cater to a wide range of developers and use cases.
GPT-4o mini has already outperformed other comparable models on various third-party benchmarks. It scored 82.0% on the Massive Multitask Language Understanding (MMLU) benchmark, surpassing Gemini Flash and Claude Haiku. Additionally, GPT-4o mini will be available through Apple Intelligence, Apple Inc.’s AI service, this fall for mobile devices and Mac desktops.
Although GPT-4o mini will initially run on OpenAI cloud servers instead of locally on devices, it still offers faster performance compared to other models. OpenAI believes that most developers are not yet interested in running models locally due to the intensive setup and computing hardware required.
GPT-4o mini will replace GPT-3.5 Turbo in ChatGPT for paying subscribers, but OpenAI will not deprecate or phase out support for GPT-3.5 Turbo in its APIs. Developers are expected to naturally migrate to GPT-4o mini due to its cost reduction and improved capabilities. Some developers, including Ramp and Superhuman, have already reported excellent results while alpha testing GPT-4o mini.
Despite the advantages of GPT-4o mini, OpenAI still recommends using the full GPT-4o model for computationally-intensive and complex applications. The higher price of GPT-4o is justified by its superior intelligence and performance in demanding tasks such as medical applications and software engineering assistance.
In conclusion, OpenAI’s release of GPT-4o mini provides developers with a more cost-efficient and capable AI model for a wide range of applications. This move solidifies OpenAI’s position as a leader in enterprise-grade AI and offers developers the opportunity to enhance their products and services with advanced language processing capabilities.