Story Content

On April 23, 2026, OpenAI released GPT-5.5, which focuses more on coding, computer use, research, and orchestration of tools. The model has rolled out to ChatGPT Plus, Pro, Business, and Enterprise users and Codex.

GPT-5.5 is especially better at code generation, with a Terminal-Bench 2.0 score of 82.7%—outperforming Anthropic Claude Opus 4.7 (69.4%) and placing it just above Anthropic Mythos Preview, the internal version of GPT-5.5. It also dominates some of the agentic and knowledge-work assessments, and it is very efficient.

OpenAI explains that GPT-5.5 is more natural in comprehending messy prompts and addressing complex workflows. Higher-level users can have a larger GPT-5.5 Pro version. The launch accelerates the AI competition with Anthropic and indicates quicker iterative processes at OpenAI.