openthoughts-tblite@2.0
Coding
OpenThoughts-TBLite: A difficulty-calibrated benchmark of 100 tasks for building terminal agents. By OpenThoughts Agent team, Snorkel AI, Bespoke Labs.
Run this task
CLI:
inspect eval inspect_harbor/openthoughts_tblite_2_0 --model openai/gpt-5Python:
from inspect_ai import eval
from inspect_harbor import openthoughts_tblite_2_0
eval(openthoughts_tblite_2_0(), model="openai/gpt-5")Dataset information
| Harbor registry | openthoughts-tblite@2.0 |
| Inspect task | openthoughts_tblite_2_0 |
| Version | 2.0 |
| Samples | 100 |
| Source | https://github.com/open-thoughts/OpenThoughts-TBLite |
See Task Parameters for the parameter set shared across all Harbor tasks.