snowflake-labs/data-eng-bench

Coding

Agentic dbt data-engineering benchmark: 103 real-world dbt tasks (analytics, dimensional modeling, incremental models, bug-fixes, snapshots) runnable hermetically on DuckDB or against Snowflake via DB_TYPE. A 30-task balanced fast subset for cost-bounded / CI runs is listed in configs/fast-30.txt.

← Back to Registry

Run this task

CLI:

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

Python:

from inspect_ai import eval
from inspect_harbor import snowflake_labs_data_eng_bench

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

Dataset information

Harbor registry	snowflake-labs/data-eng-bench
Inspect task	`snowflake_labs_data_eng_bench`
Latest digest	sha256:de31878dba7e5a9aad85dc8276820b0a20ab593aed46ff93dfbbbdb6007ec995
Samples	103
Source	https://github.com/Snowflake-Labs/data-eng-bench

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