Even advanced language models like GPT-5.2 struggle with basic counting tasks, highlighting fundamental limitations in achieving error-free AI systems.

Current large language models, including GPT-5.2, fail at simple arithmetic operations like counting to five

The research reveals a gap between perceived AI capabilities and actual reliable performance on elementary tasks

Study suggests significant challenges remain in developing trustworthy LLMs that can guarantee zero-error performance on basic operations

Findings raise questions about the reliability of large language models for tasks requiring precise numerical accuracy