- OpenAI has launched the GPT-4o mini model, its fastest and cheapest model that scores 82% on the MMLU benchmark.
- GPT-4o mini is even cheaper than Gemini 1.5 Flash and Claude 3 Haiku. It performs significantly better than other competing models in its class.
- GPT-4o mini is already live on ChatGPT for free and paid users. Developers can also start using the model through the API.
After two months of releasing the GPT-4o (‘Omni’) model, OpenAI has introduced a faster and cheaper model — GPT-4o mini. The mini model is lightweight and small yet it scores 82% on the MMLU benchmark which is remarkable. For context, the flagship GPT-4o model scored 88.7 on the same MMLU benchmark test.
Apart from its capability, the GPT-4o mini model is incredibly cheap. It’s even more affordable than Google’s Gemini 1.5 Flash and Anthropic’s Claude 3 Haiku models. It costs $0.15 to process 1 million input tokens and $0.60 for 1 million output tokens. On top of that, it’s a multimodal model which means it can reason with images and audio files too.
As for general users, GPT-4o mini is already live for all ChatGPT users. Even if you have a free ChatGPT account, you can access the mini model right away. Moreover, ChatGPT Plus and Team users can also access the new OpenAI model starting today.
By the way, on the LMSYS leaderboard, the GPT-4o mini model outperforms Anthropic’s largest Claude 3 Opus and OpenAI’s GPT-4 models. It’s evident that OpenAI has again built a strong model that is fast, cheap, and can be used by developers for all kinds of applications.
Google and Anthropic made headlines after releasing their cheapest and fastest Gemini Flash and Haiku models. But again, OpenAI has one-upped the competition with a solid lightweight model that doesn’t compromise on performance. Developers are likely to move to the new GPT-4o mini model from GPT-3.5 Turbo, Gemini Flash, and Claude Haiku.
