OpenAI Releases GPT-5.5, a Fully Retrained Agentic Model That Scores 82.7% on Terminal-Bench 2.0 and 84.9% on GDPval

Editor's take

OpenAI has launched GPT-5.5, a significantly retrained language model designed for autonomous task execution across coding, research, data analysis, and software operation. This development moves beyond sequential instruction following, aiming for a more integrated and self-directed approach to complex computational workflows.

The significance lies in GPT-5.5's potential to automate a broader spectrum of professional tasks, reducing the need for human oversight in multi-step processes. Its performance on benchmarks like Terminal-Bench 2.0 (82.7%) and GDPval (84.9%) suggests a tangible advancement in agentic capabilities, potentially impacting productivity tools and the future of software development.

Future developments to monitor include the model's actual deployment in real-world enterprise settings and its performance on more dynamic, less structured tasks. The economics of running such an agentic model at scale, and the emergence of specialized agents built upon this foundation, will also be critical indicators of its long-term impact.