GPT-5.5 is OpenAI’s most capable agentic AI model yet

OpenAI launched the GPT-5.5 model on April 23, describing it as a "new class of intelligence for real work." This model is designed to excel in planning, tool usage, and output verification, enhancing performance autonomy. GPT-5.5 is the first retrained core model since GPT-4.5, developed in collaboration with NVIDIA's GB200 and GB300 systems. The practical difference is that tasks requiring multiple human interventions can now be fully delegated to the model. Results show GPT-5.5 achieved 82.7% on the Terminal-Bench 2.0 benchmark, outperforming GPT-5.4 and Claude Opus 4.7.