OpenAI’s AI model GPT-4, known from ChatGPT, was the leader for almost a year in an independent ranking with which researchers compare the performance of AI models. But GPT-4 has now lost its leading position: there is a competitor with an even better AI model.
The Chatbot Arena ranking, which compares the performance of AI language models, has a new number 1. GPT-4 Turbo has been overtaken by Claude 3 Opus from the startup Anthropic.
It sparked a flood of responses from AI experts on social media. Understandable, because for almost a year several variations of GPT-4 topped Chatbot Arena’s list. The fact that OpenAI has now been surpassed is seen as a remarkable changing of the guard.
Claude
Also notable is the relatively good performance of another AI model from Anthropic, Claude 3 Haiku. According to the startup, this is the ‘fastest and most affordable model in its segment’. The American company launched its Claude 3 models at the beginning of this month. The new Claude models are available to users with a Claude Pro subscription, which costs $20 per month.
“For the first time, the best available models – Opus for advanced tasks, Haiku for cost and efficiency – come from a company other than OpenAI,” AI researcher Simon Willison told Ars Technica. He calls this ‘reassuring’ because it means that there is more competition at the top of the AI suppliers.
GPT-5 is coming
GPT-4 has been around for more than a year. GPT-5 is expected this summer and that version should be ‘substantially better’ than its predecessor. OpenAI is now also busy offering its AI services to Hollywood.
Chatbot Arena is run by a research organization and collaborates with three American universities. The rankings are created based on crowdsourced reviews.
Also watch our video in which Bright presenter Eric completely replaces himself with AI.
Read more news about AI and stay informed with our newsletter.
