terminal-bench-science/terminal-bench-science

Science
Coding

A benchmark for evaluating AI agents on research workflows across the life, physical, earth, mathematical, and engineering sciences.

← Back to Registry

Run this task

CLI:

inspect eval inspect_harbor/terminal_bench_science --model openai/gpt-5

Python:

from inspect_ai import eval
from inspect_harbor import terminal_bench_science

eval(terminal_bench_science(), model="openai/gpt-5")

Dataset information

Harbor registry terminal-bench-science/terminal-bench-science
Inspect task terminal_bench_science
Latest digest sha256:91531bf50016a7c64f6cc60794a17c64c6b2c14858a8ae0de39ca16f2abd611a
Samples 70
Source https://github.com/harbor-framework/terminal-bench-science

See Task Parameters for the parameter set shared across all Harbor tasks.