OpenAI simply introduced GPT-4o mini – a brand new compact however highly effective AI mannequin aiming to make the GPT-4 engine extra inexpensive.
The worth is ready at 15 cents per 1 million enter tokens and 60 cents per 1 million output tokens, which is quite a bit cheaper than GPT-3.5 Turbo.
OpenAI revealed that 4o mini is far more highly effective than different fashions equivalent to Gemini Flash, Claude Haiku, and GPT-3.5 Turbo. It surpasses all of them in benchmarks and produces outcomes which can be practically pretty much as good as GPT-4o.
The small mannequin is geared toward textual intelligence and multimodal reasoning, the corporate revealed in a press launch. GPT-4o mini is best than different small fashions at reasoning duties, excels in mathematical reasoning and coding, and even performes impressively for multimodal reasoning.
The engine has a context window of 128k tokens and has information as much as October 2023. Picture, video, and audio help is coming sooner or later.
Implementation of GPT-4o mini is beginning right now, and it’s changing GPT 3.5 in Free, Plus, and Workforce subscriptions for finish customers. Enterprise shoppers could have entry to the mannequin beginning subsequent week.
Supply