89%
89 of 100 right on its latest run, Sep 26. Its first verdict comes on Oct 2, 2026.
Runs so far2
Verdict fromOct 2
Baseline so far: Sep 25–26, 2026 (2 runs, 200 answers). Released Sep 2, so the first 22 days aren't measured.
Score, every run
Share of questions answered correctly. Grey band: the baseline's normal range. Hover for a run.
80%90%100%Sep 25Sep 26
Sep 2689%
Sep 2589%
By question type
All runs so farAnswersRight
Reasoning4098%
Logic puzzles4093%
Reading code4088%
Following instructions4095%
Long documents4073%
Raw numbers
Per runAverage so farLatest run
Refused to answer0.00.0
Input tokens per question4,4174,417
Output tokens per question3,5283,579
Thinking tokens per question3,3573,417
Cost of 100 questions$2.03$2.07