OpenAI just made ChatGPT a much better writer. The latest update to the underlying GPT-4o model gave the AI a boost in creativity.

This new update also helped it leapfrog Google Gemini and reclaim the top spot in the LLM arena. This is a virtual venue where users rank the output of two models without knowing which is which until the end.

OpenAI didn't share much about the new update other than that its "creative writing ability has leveled up" and that it is now "more natural and engaging" with "more tailored writing to improve relevance and readability".

OpenAI has been on its game recently, adding a range of new features including bringing Advanced Voice to the web and hints at better DALL-E image generations. There are even rumors circulating of a version of Sora coming soon to bring AI video to the rapidly evolving AI platform.

The latest update is more 'behind the scenes' than a new UI or flashy new features like the impressive Canvas. It is a change to how GPT-4o works, making it more of a creativity powerhouse than previous generations.

The new version of GPT-4o, still called GPT-4o, is also better at working with files you've uploaded to ChatGPT and providing deeper insights into its contents.

It was released in secret to the lmarena.ai (formerly LMSys arena) LLM chatbot arena. This is a platform where models compete anonymously to find out which ones score better with human users.

OpenAI dropped it in as "anonymous-chatbot" last week and it quickly surpassed Gemini-Ex-1114, the latest model from Google that held the crown for about a week. Writing on X, Lmarena described it as a "remarkable improvement" on previous versions including around creative writing, coding and math.

Have you noticed a change in ChatGPT over the past few days? I've tried it out and it does seem to be more complete and engaging in its responses. This upgrade in creativity might also explain why DALL-E images have improved.