OpenAI officially launched its latest artificial intelligence model, GPT-5.5, on April 23, 2026, making it available to paid subscribers of ChatGPT and Codex. The company touts the new model as its "smartest and most intuitive" offering to date, designed to significantly enhance coding, computer use, and research capabilities. Early user 'nic' quickly demonstrated its creative potential, tweeting, > "gpt-5.5 in chatgpt made me a theme for my chatgpt," highlighting the model's immediate impact on user experience and customization.
GPT-5.5 introduces substantial advancements, particularly in agentic coding and complex knowledge work. OpenAI reported that the model achieved a state-of-the-art accuracy of 82.7% on Terminal-Bench 2.0, a benchmark designed for evaluating complex command-line workflows. Furthermore, it demonstrated strong performance on SWE-Bench Pro, successfully resolving 58.6% of real-world GitHub issues end-to-end in a single pass. These metrics underscore its enhanced ability to plan, utilize tools, and navigate ambiguity in intricate tasks.
The new model is currently rolling out to Plus, Pro, Business, and Enterprise users within ChatGPT and its dedicated coding assistant, Codex. OpenAI stated that GPT-5.5 excels at analyzing data, writing and debugging code, operating software, researching online, and creating documents and spreadsheets. This broad utility aims to streamline complex workflows and enable a new, more integrated way of interacting with computers. The company also confirmed that API deployments for GPT-5.5 and GPT-5.5 Pro are expected "very soon," broadening its accessibility for developers and enterprises.
OpenAI emphasized the robust safeguards integrated into GPT-5.5, noting it underwent extensive third-party testing and red-teaming for cyber and bio risks. While classified as "High" risk due to its advanced capabilities, the model did not cross the "Critical" cybersecurity threshold, which could lead to "unprecedented new pathways to severe harm." This rapid release, following GPT-5.4 by less than two months, underscores the intense competitive pressure in the AI sector, with OpenAI actively racing rivals like Google and Anthropic to deliver increasingly capable and secure models.
