File tree Expand file tree Collapse file tree
Expand file tree Collapse file tree Original file line number Diff line number Diff line change @@ -58,13 +58,13 @@ models:
5858# Per-sample spend cap -- the real budget. Requires model_cost_config below for
5959# every model used, or Hawk/Inspect errors out.
6060#
61- # COST (2 models x 68 samples x 1 epoch = 136 samples, $100 /sample cap):
62- # Max: 136 x $100 = $13,600 hard ceiling -- every sample hits its cap
61+ # COST (2 models x 68 samples x 1 epoch = 136 samples, $200 /sample cap):
62+ # Max: 136 x $200 = $27,200 hard ceiling -- every sample hits its cap
6363# Expected sits near max: runs are gated (default), so an unsolved problem
6464# retries until it hits the cap, and the expected solve rate on this selection
6565# is ~0% for current models.
66- cost_limit : 100 .0
67- working_limit : 129_600 # 36 * 60 * 60
66+ cost_limit : 200 .0
67+ working_limit : 259_200 # 72 * 60 * 60
6868
6969# Prices in dollars per 1M tokens (from scripts/data/model_prices_and_context_window.json).
7070# Keys must match Inspect's resolved <provider>/<model> names.
You can’t perform that action at this time.
0 commit comments