GPT-4o Mini, said OpenAI, seeks to make the AI technology of the startup “much more affordable, climate-friendly, and available to everyone.” Starting July 18, the model became available to ChatGPT-free users buying Plus and Team subscriptions. Starting next week, enterprise users will have access, according to the statement.

OpenAI CEO Sam Altman claims that GPT-4o mini is more than 60% less expensive than GPT-3.5 Turbo since its cost is 15 cents per million input tokens and 60 cents per million output tokens. The new device, he explained, costs one hundred times less than the Da Vinci 003.

Altman characterized the ChatGPT-4o mini in an X post—formerly Twitter—as “technology too cheap to meter.”

Related: ChatGPT Edu: Enhancing University Education with AI

OpenAI said in a blog post that the GPT-4o mini scores 82% on Mass Multitask Language Understanding (MMLU) and performs better than the GPT-4 model on chat preferences. MMLU gauges language models’ text intelligence and reasoning capacity. A higher score indicates that one can use language effectively in many spheres of real life.

Google’s Gemini Flash came in at 77.9% on the MMLU test and Anthropic’s Claude Haiku at 73.8%, OpenAI reported. Having knowledge up to October 2023, GPT-4o mini also outperformed its nearest rivals in arithmetic and coding challenges.

According to OpenAI, the mini model incorporates fresh safety mechanisms to withstand so-called quick injections and jailbreaks. It clarified that the safety elements raise the dependability of the answers in the model and so enhance safety in major applications.

Currently accessible in the API, or application programming interface, GPT-4o mini is a text and vision model. Future support for text, image, video, and audio inputs and outputs will be provided.

You can also freely share your thoughts and comments about the topic in the comment section. Additionally, don’t forget to follow us on our Telegram, YouTube, and Twitter channels for the latest news and updates.

The post GPT-4o Mini: OpenAI’s Climate-Friendly AI Model appeared first on Coin Engineer.