Terminal-Bench 3.0

Label Shift MonitorML & Forecasting

Human baseline3h

TaxonomyML & ForecastingML & Forecasting

All trialsGemini-3gemini-cliTrial 01

0%

Trial reward

Gemini-3Fail

Trial 01 — 8th of 8 by reward score. This model averages 87.5% across 8 trials on this task.

  • $0.69Cost
  • 3.4mTime
  • 19kOutput tokens

Failed — graded by execution verdict, the complete grading on this bench.