On the LLM comparison platform Chatbot Arena, Claude 3 Opus surpassed GPT-4 for the first time. Is a transition of power underway?

“The king is dead,” says X. GPT-4 is no longer number one on the LLM platform Chatbot Arena and has to give up its place to Claude 3 Opus. This open source platform allows you to compare LLMs one by one to compare the answers they give to the same question. Users then vote on which model they think is the best: Chatbot Arena can be viewed as a “hit list” for LLM chatbots.

Magnus Opus

Since the launch of GPT-4 in spring 2023, OpenAI’s LLM has consistently been at the top of the rankings. But since March 26th, a new number one has been on display: Claude 3 Opus, the new flagship model from Anthropic. Claude 3 Opus launched in March and at launch Anthropic showed benchmarks that showed the model was superior to GPT-4 and Google Gemini. These words don’t seem to be a lie.

There are still reasons for Anthropic to print out the current rankings and display them in office buildings. The smaller, faster model Claude 3 Haiku is also very popular and takes sixth place. Claude Sonnet is tied for fourth place with Google Gemini Pro.

Change of power?

Chatbot Arena is a popular tool for researchers to compare the skills of LLMs. So it means that Claude 3 is rated higher than GPT-4. OpenAI’s LLMs have long been considered a reference for generative AI, but Anthropic is increasingly emerging as a formidable opponent.

OpenAI also understood this message. We are working hard on a new version of GPT. It’s still unclear when exactly it will come into the world, but rumors that it will happen this year are getting louder. It is also still unclear whether the new version GPT-4.5 will be called GPT-5.