primer/agent-eval

Agent eval navigation

005-agent-enables-theme-switching

Recommended results are shown first, followed by the percent change from Control in parentheses. Positive means the raw value increased and negative means it decreased. Models are ranked by test pass rate within each scenario.

ModelEffortTests passedTurnsOutput tokensPremium requestsAPI timeSession timeTool calls
gpt-5.6-sol0/18 (0%)11 (+57.1%)10,265 (+42%)0 (0%)135.7s (+52.3%)153.4s (+58.5%)24 (+60%)
gpt-5.6-terra0/18 (0%)11 (+22.2%)8,466 (+123.1%)0 (0%)79.6s (+153.1%)90.1s (+128.1%)20 (+17.6%)
gemini-3.1-pro-preview0/18 (0%)25 (+92.3%)1,664 (+138.7%)1 (0%)100.3s (+133.7%)126.8s (+143.6%)26 (+73.3%)
claude-sonnet-50/18 (0%)28 (+16.7%)7,565 (+9.5%)1 (0%)97.6s (+17.1%)118.7s (+21.1%)30 (+15.4%)
claude-opus-50/18 (0%)49 (+104.2%)25,957 (+149.2%)1 (0%)321.6s (+156.7%)463.8s (+214.5%)56 (+124%)
gemini-3.6-flash0/18 (0%)75 (+82.9%)7,503 (-15.4%)14 (0%)515.9s (+35.6%)619.7s (+34.2%)79 (+92.7%)

Strong lines show Recommended results and muted lines show Control over time. The comparison table shows each Recommended result relative to Control.

Tests passed

19.89.9007-2908-0308-1008-17Run date

Turns

136.968.5007-2908-0308-1008-17Run date

Output tokens

38,351.119,175.6007-2908-0308-1008-17Run date

Premium requests

15.47.7007-2908-0308-1008-17Run date

API time

564.5282.2007-2908-0308-1008-17Run date

Session time

677.7338.8007-2908-0308-1008-17Run date

Tool calls

13670407-2908-0308-1008-17Run date

Change from Control

Recommended results are shown first, followed by the percent change from Control in parentheses.

Model
claude-opus-50/18 (0%)1/18 ()15/18 ()0/18 (0%)
claude-sonnet-50/18 (0%)0/18 (0%)0/18 (0%)0/18 (0%)
gemini-3.1-pro-preview14/18 ()14/18 ()14/18 ()0/18 (0%)
gemini-3.6-flash0/18 (0%)0/18 (-100%)0/18 (0%)0/18 (0%)
gpt-5.6-sol1/18 ()16/18 ()1/18 ()0/18 (0%)
gpt-5.6-terra0/18 (0%)0/18 (0%)0/18 (0%)0/18 (0%)
View raw trend data
DateModelTests passedTurnsOutput tokensPremium requestsAPI timeSession timeTool calls
claude-opus-50/18 (0%)30 (+7.1%)15,853 (+23.7%)1 (0%)200.2s (+20%)226.2s (+19%)32 (0%)
claude-sonnet-50/18 (0%)30 (+50%)8,532 (+7.8%)1 (0%)130.1s (+29.9%)154s (+32.3%)32 (+45.5%)
gemini-3.1-pro-preview14/18 ()14 (-44%)1,008 (-30%)1 (0%)72.6s (-32.6%)88.5s (-29.1%)15 (-40%)
gemini-3.6-flash0/18 (0%)104 (+205.9%)19,265 (+269.6%)14 (0%)325.1s (+146.8%)442.8s (+178.2%)110 (+223.5%)
gpt-5.6-sol1/18 ()23 (+283.3%)14,215 (+102.6%)0 (0%)296.7s (+126.8%)327.1s (+133.7%)51 (+218.8%)
gpt-5.6-terra0/18 (0%)10 (+25%)3,019 (-48.4%)0 (0%)43.5s (-9.6%)65.3s (+16.2%)22 (+37.5%)
claude-opus-51/18 ()80 (+300%)34,928 (+240.7%)1 (0%)464.3s (+281.7%)569.3s (+290.6%)83 (+219.2%)
claude-sonnet-50/18 (0%)31 (+10.7%)9,368 (-2%)1 (0%)128.4s (+6.2%)150.8s (+7.9%)33 (+10%)
gemini-3.1-pro-preview14/18 ()33 (+32%)3,579 (+146.1%)1 (0%)132.4s (+64.4%)172.4s (+72.3%)36 (+38.5%)
gemini-3.6-flash0/18 (-100%)65 (+14%)18,000 (+306.9%)14 (0%)202.3s (+59.2%)309.6s (-40.8%)71 (+24.6%)
gpt-5.6-sol16/18 ()26 (+271.4%)13,065 (+192.2%)0 (0%)298.1s (+251.8%)332.7s (+259.3%)72 (+300%)
gpt-5.6-terra0/18 (0%)16 (+45.5%)9,250 (+134.3%)0 (0%)81.4s (+171%)108.7s (+104.5%)31 (+63.2%)
claude-opus-515/18 ()48 (+100%)18,445 (+52.7%)1 (0%)261.5s (+47.4%)311.2s (+59.1%)49 (+96%)
claude-sonnet-50/18 (0%)32 (+18.5%)10,176 (+47.3%)1 (0%)116.4s (+8.9%)141.1s (-4.1%)34 (+21.4%)
gemini-3.1-pro-preview14/18 ()17 (-19%)1,083 (+9.6%)1 (0%)56.2s (-5.9%)70.8s (-1.8%)17 (-19%)
gemini-3.6-flash0/18 (0%)125 (+78.6%)33,077 (+113%)14 (0%)345.8s (+113.4%)480.5s (+82.3%)125 (+78.6%)
gpt-5.6-sol1/18 ()25 (+257.1%)7,326 (+113.3%)0 (0%)104.4s (+138.5%)126.7s (+141.3%)45 (+164.7%)
gpt-5.6-terra0/18 (0%)8 (-20%)4,629 (-19.6%)0 (0%)40.3s (+5.8%)53.4s (+15.3%)21 (+10.5%)
claude-opus-50/18 (0%)49 (+104.2%)25,957 (+149.2%)1 (0%)321.6s (+156.7%)463.8s (+214.5%)56 (+124%)
claude-sonnet-50/18 (0%)28 (+16.7%)7,565 (+9.5%)1 (0%)97.6s (+17.1%)118.7s (+21.1%)30 (+15.4%)
gemini-3.1-pro-preview0/18 (0%)25 (+92.3%)1,664 (+138.7%)1 (0%)100.3s (+133.7%)126.8s (+143.6%)26 (+73.3%)
gemini-3.6-flash0/18 (0%)75 (+82.9%)7,503 (-15.4%)14 (0%)515.9s (+35.6%)619.7s (+34.2%)79 (+92.7%)
gpt-5.6-sol0/18 (0%)11 (+57.1%)10,265 (+42%)0 (0%)135.7s (+52.3%)153.4s (+58.5%)24 (+60%)
gpt-5.6-terra0/18 (0%)11 (+22.2%)8,466 (+123.1%)0 (0%)79.6s (+153.1%)90.1s (+128.1%)20 (+17.6%)