During its Spring Update event in May, OpenAI announced its new large language model (LLM), GPT-4o. GPT-4o is a state-of-the-art LLM that leads in various industry benchmarks, but it comes at a cost. GPT-4o is one of the most expensive models available; only Anthropic"s Claude 3 Opus costs more.
Over the past few months, all leading AI companies have been releasing cost-effective AI models to lower costs for developers. Claude 3 Haiku and Gemini 1.5 Flash are examples of such models. Today, OpenAI announced its own cost-effective AI model called GPT-4o mini. The new GPT-4o mini model will be cheaper and faster than the current OpenAI models, and it will replace the GPT-3.5 Turbo model in OpenAI"s lineup.
GPT-4o mini has achieved an 82% score on MMLU, which is decent compared to GPT-4o"s score of 89%. The main competitors to GPT-4o mini are the Gemini 1.5 Flash and Claude 3 Haiku models, which have scored 79% and 75% on MMLU, respectively.
GPT-4o mini will cost approximately $0.60 USD per million output tokens and $0.15 USD per million input tokens. This is 60% cheaper than the GPT-3.5 Turbo model it will replace. It also matches the price of Gemini 1.5 Flash and Claude 3 Haiku. The context window remains 128,000 tokens, and it comes with a knowledge cutoff of October 2023.
The new GPT-4o mini model is now available for developers via APIs. Consumers can also use this model through ChatGPT web and mobile apps. Enterprise users will gain access to GPT-4o mini next week.
With the release of GPT-4o mini, OpenAI has taken a significant step towards making advanced AI more accessible and affordable for a wider range of users. While maintaining a competitive edge in performance, this new model addresses the growing demand for cost-effective solutions in the AI landscape.
via: CNBC