frontier-bench/frontier-bench
Coding
Assistants
Frontier-Bench: a digest-pinned suite of hard agentic terminal tasks.
Run this task
CLI:
inspect eval inspect_harbor/frontier_bench --model openai/gpt-5Python:
from inspect_ai import eval
from inspect_harbor import frontier_bench
eval(frontier_bench(), model="openai/gpt-5")Dataset information
| Harbor registry | frontier-bench/frontier-bench |
| Inspect task | frontier_bench |
| Latest digest | sha256:97fd2ba3aabdda16823a1a8ea695a3875e50e800caa60b450686deedc7171763 |
| Samples | 74 |
| Source | https://github.com/harbor-framework/frontier-bench |
See Task Parameters for the parameter set shared across all Harbor tasks.