OpenAI Launches GPT-4o Mini, A Smaller And Cheaper Version

OpenAI has launched a smaller and cheaper version of its flagship GPT-4o for web and mobile app consumers, including developers.

OpenAI has launched GPT-4o mini, a smaller and cheaper version of GPT-4o. The small AI model outperforms existing cutting-edge AI models. The mini version was released yesterday for developers, web, and mobile application users of ChatGPT. The AI model will be released for Enterprise users by next week.

For tasks that involve text and vision reasoning, GPT-4o mini outperforms Gemini 1.5 Flash, Llama 3 (70B), NeMo, GPT-3.5 Turbo, Reka Edge, and many more competitive smaller models. An independent Artificial Analysis shows a glaring difference in the performance results between GPT-4o and other available small AI models. GPT-40 mini scored 82% on MMLU, a benchmark to measure reasoning, compared to 79% for Gemini 1.5 Flash and 75% for Claude 3 Haiku.

“For every corner of the world to be empowered by AI, we need to make the models much more affordable. I think GPT-4o mini is a really big step forward in that direction”

- OpenAI’s head of Product API

GPT-4o mini will also replace GPT-3.5 Turbo as the smallest AI model offered by the company. It is still unclear if the existing users of Turbo will be shifted to GPT-4o mini or not. The company claims that its offering is much more affordable and consumes less power. Mini comes with text and vision capabilities, but the company says that video and audio capabilities will be added soon.

“Relative to comparable models, GPT-4o mini is very fast, with a median output speed of 202 tokens per second. This is more than 2X faster than GPT-4o and GPT-3.5 Turbo and represents a compelling offering for speed-dependent use-cases including many consumer applications and agentic approaches to using LLMs,” said George Cameron, Co-Founder at Artificial Analysis, in an email to a tech media house.

For developers, GPT-4o mini is priced at 15 cents per million input tokens and 60 cents per million output tokens. The model comes with a context window of 128,000 tokens that transforms roughly to the length of a book. OpenAI has not revealed how big the GPT-4o mini actually is but claims that it is similar to other small AI models.