OpenAI GPT-5.5 takes ChatGPT back ahead of Claude

After launching its answer to Nano Banana, OpenAI has introduced GPT-5.5. The upgrades focus on productivity, with the aim of helping ChatGPT move back ahead of Claude.

Although ChatGPT remains number one in the AI market, rival services have gained ground in recent months. OpenAI is now facing a particular threat from Anthropic, whose product development has concentrated on productivity improvements. With GPT-5.5, however, OpenAI intends to regain the upper hand. The model is being rolled out to paid versions of ChatGPT and promises substantial time savings for developers as well as employees handling administrative work. OpenAI describes it as a “new form of intelligence for real work”.

GPT-5.5 productivity improvements for real work

GPT-5.5 is smarter than earlier models and has a better grasp of what users expect. As a result, it can take on a larger share of the work involved in a particular task. The new AI can handle more complicated assignments, plan its own actions, use tools when required and check its own output. The benefits are said to be especially significant for coding, when users ask the AI to control a computer, and for everyday office tasks.

When used in the Codex application, GPT-5.5 also outperforms its predecessor when asked to create documents such as spreadsheets or presentations. Within OpenAI, GPT-5.5 is reportedly already used by numerous departments, including communications and finance. Its finance teams, for instance, used the tool to analyse 71,637 pages of tax documents. A member of the “go-to-market” team is also said to have used Codex and GPT-5.5 to produce sales reports, saving 5 to 10 working hours each week.

OpenAI catches up with its competitors

In any case, GPT-5.5 allows OpenAI to catch up with competing models. In Artificial Analysis’s rankings, GPT-5.4 (xhigh) sits behind Gemini 3.1 Pro and Claude Opus 4.7, Anthropic’s recently released AI. GPT-5.5, however, ranks above both competitors and becomes the best AI according to this league table. Artificial Analysis ranks models using the results of several assessments. It also says that GPT-5.5 holds the number-one position on GDPval-AA, an evaluation that tests AI capabilities on real-world tasks across a range of professions.

This week, OpenAI also released a new image generator for ChatGPT which, according to Arena.ai, performs better than Nano Banana.

What we think

OpenAI has plainly reassessed its priorities to avoid falling behind Claude and Gemini. With GPT-5.5, it has delivered a rival to Claude Opus 4.7, whose key strength is productivity. The new image generator also produces better results than Nano Banana, the technology that helped Gemini create a buzz in 2025.