All terms
Evaluation
Terminal-Bench
A benchmark that measures how well AI agents complete real tasks inside a command-line terminal.
Definition
Terminal-Bench is a benchmark that evaluates AI agents on their ability to get things done in a command-line terminal — installing software, editing files, running programs, and fixing what breaks — across a set of realistic, often multi-step tasks. Because a terminal exposes the full power of a computer, doing well requires planning, correct tool use, and error recovery over many steps, making it a demanding test of agentic coding. It is widely cited when labs compare models on autonomous coding, and it is the predecessor to Frontier-Bench.