OpenAI Reveals New Small AI Model GPT-4o Mini
- The GPT-4o mini model is designed to be 60% more cost-efficient than GPT-3.5 Turbo.
- The GPT-4o mini outperforms existing models, scoring 82% on the Massive Multitask Language.
- Starting Thursday, the GPT-4o mini will be available to free ChatGPT users, as well as ChatGPT Plus and Team subscribers.
OpenAI GPT-4o mini: Smaller, Smarter & Cheaper AI Model
OpenAI has revealed the GPT-4o mini model, a more intelligent and affordable alternative to existing AI models. Developed to be both cost and energy-efficient, this new model aims to make advanced AI technology accessible to a broader range of customers, including small-scale companies and those with limited resources.
Decoding the GPT-4o mini Model
OpenAI launched the GPT-4o mini on Thursday, the company said that it is more intelligent and affordable compared to the existing models. OpenAI said the main objective of the new AI model is to make it more affordable and energy efficient and spread the AI technology to a broader range of customers.
Moreover, the new GPT-4o mini is 60% cost efficient than GPT-3.5 Turbo as it is priced at 15 cents per million input tokens and 60 cents per million output tokens.
It also surpassed the GPT-4 model regarding chat preferences and scored 82% on Massive Multitask Language Understanding (MMLU).
The Massive Multitask Language Understanding (MMLU) benchmark assesses the textual intelligence and reasoning abilities of language models. A higher MMLU score indicates superior comprehension and application of language across diverse fields, which translates to improved performance in real-world scenarios.
The company also compared the GPT-4o mini model's score with Google's Gemini Flash and Anthropic's Claude Haiku. Google's Gemini Flash score was 77.9% and Anthropic's Claude Haiku was 73.8%.
The GPT-4o mini model uses a minimal amount of computational power making it more affordable and accessible to small-scale companies. This mini model is also helpful for companies with fewer resources that want to integrate AI technology into their operations.
The new AI model currently supports text and vision in its API, the company will eventually expand to include text, image, video, and audio inputs and outputs in the future.
According to their press release, GPT-4o mini will be accessible to free ChatGPT users, as well as ChatGPT Plus and Team subscribers, starting Thursday. Enterprise users will gain access to the mini model next week.
Conclusion
With its impressive performance metrics, including a 60% cost efficiency over GPT-3.5 Turbo and an 82% score on the Massive Multitask Language Understanding (MMLU) benchmark, GPT-4o mini sets a new standard in AI technology. As it expands to support a wider range of inputs and outputs, this model promises to revolutionize AI integration for businesses of all sizes. Starting Thursday, ChatGPT users across various subscription tiers will gain access to this innovative model, with enterprise users following next week.
