OpenRouter · controlled comparisonExploratory benchmark

Jev / Luna

How much quality, how quickly, at what cost. Identical output schemas; separately scored tasks with checkable answers.

Measured requests3603 distinct experiments
Valid responses360100.0% schema success
Known total billing$0.0115500 missing bills · includes warmups
Warmup requests4$0.000087 known billing · excluded from scores

Quality and response-time tables use valid measured responses. “Accuracy · all” also counts invalid requests as failures. Cost includes all measured requests. A dash means unavailable; unknown cost is never assumed to be zero.

Observed labels

Reading comprehension · BoolQ

60 paired cases

Accuracy: higher is better. Brier and log loss: lower is better.

ModelValid / requestsAccuracy · validAccuracy · allBrierLog lossp50 / p95USD / 1,000USD / correct
Jev60 / 6095.0%95.0%0.0680.2480.236 / 0.372 s$0.019847$0.000021
Luna60 / 6081.7%81.7%0.1683.6761.148 / 1.470 s$0.044483$0.000054

Median complete-response time

Jev
0.236 s
Luna
1.148 s

Valid responses only. Smaller is faster.

Paired comparison · 95% intervals

Luna − Jev accuracy-13.33 pp [-21.67, -5.00]

Luna / Jev median time4.86× [4.42, 5.10]

60 of 60 cases paired; 0 excluded. Positive accuracy difference favours Luna. A time ratio below 1 favours Luna.
Billing and routing detail
  • Jev: $0.001191 known measured billing · 0 missing bills · resolved model typesafe/jev-1.13-20260917 · provider TypeSafe
  • Luna: $0.002669 known measured billing · 0 missing bills · resolved model openai/gpt-6-luna · provider OpenAI

All-request timing, additional paired metrics, interval seeds, and exact unrounded numbers are in summary.json. Billing includes invalid measured requests; warmups are shown separately above.

Observed labels

Policy decisions

60 paired cases

Accuracy: higher is better. Brier and log loss: lower is better.

ModelValid / requestsAccuracy · validAccuracy · allBrierLog lossp50 / p95USD / 1,000USD / correct
Jev60 / 60100.0%100.0%0.0080.0670.243 / 0.331 s$0.020927$0.000021
Luna60 / 60100.0%100.0%0.0000.0001.149 / 1.647 s$0.045735$0.000046

Median complete-response time

Jev
0.243 s
Luna
1.149 s

Valid responses only. Smaller is faster.

Paired comparison · 95% intervals

Luna − Jev accuracy+0.00 pp [0.00, 0.00]

Luna / Jev median time4.72× [4.26, 5.05]

60 of 60 cases paired; 0 excluded. Positive accuracy difference favours Luna. A time ratio below 1 favours Luna.
Billing and routing detail
  • Jev: $0.001256 known measured billing · 0 missing bills · resolved model typesafe/jev-1.13-20260917 · provider TypeSafe
  • Luna: $0.002744 known measured billing · 0 missing bills · resolved model openai/gpt-6-luna · provider OpenAI

All-request timing, additional paired metrics, interval seeds, and exact unrounded numbers are in summary.json. Billing includes invalid measured requests; warmups are shown separately above.

Known mathematical probabilities

Exact probabilities

60 paired cases

MAE, RMSE, and excess Brier: lower is better. Expected Brier includes irreducible outcome uncertainty.

ModelValid / requestsMAERMSEExcess BrierExpected Brierp50 / p95USD / 1,000
Jev60 / 600.0640.1000.0100.1560.237 / 0.311 s$0.018645
Luna60 / 600.0330.0740.0060.1521.143 / 1.596 s$0.041408

Median complete-response time

Jev
0.237 s
Luna
1.143 s

Valid responses only. Smaller is faster.

Paired comparison · 95% intervals

Luna − Jev mae-0.03 [-0.05, -0.01]

Luna / Jev median time4.81× [4.47, 5.37]

60 of 60 cases paired; 0 excluded. Negative MAE difference favours Luna. A time ratio below 1 favours Luna.
Billing and routing detail
  • Jev: $0.001119 known measured billing · 0 missing bills · resolved model typesafe/jev-1.13-20260917 · provider TypeSafe
  • Luna: $0.002485 known measured billing · 0 missing bills · resolved model openai/gpt-6-luna · provider OpenAI

All-request timing, additional paired metrics, interval seeds, and exact unrounded numbers are in summary.json. Billing includes invalid measured requests; warmups are shown separately above.

Read the result in context

Every request

Inspect the record behind each aggregate. Includes warmups and failed requests.

364 requests shown

CaseExperimentModelPhaseTargetp(yes)ResultSecondsUSDError
warmup-yeswarmupLunawarmup1.00001.0000Valid1.134$0.000030
warmup-yeswarmupJevwarmup1.00000.9600Valid0.241$0.000014
warmup-nowarmupJevwarmup0.00000.0400Valid0.251$0.000014
warmup-nowarmupLunawarmup0.00000.0000Valid1.973$0.000030
policy-016policyLunameasured0.00000.0000Valid0.900$0.000046
policy-016policyJevmeasured0.00000.0600Valid0.238$0.000021
probability-044probabilityJevmeasured0.06670.1900Valid0.219$0.000017
probability-044probabilityLunameasured0.06670.0667Valid1.040$0.000041
boolq-dev-1550boolqLunameasured0.00000.0000Valid1.329$0.000055
boolq-dev-1550boolqJevmeasured0.00000.0600Valid0.347$0.000024
boolq-dev-0579boolqJevmeasured1.00000.6300Valid0.240$0.000018
boolq-dev-0579boolqLunameasured1.00001.0000Valid1.316$0.000040
boolq-dev-1066boolqJevmeasured1.00000.8600Valid0.231$0.000020
boolq-dev-1066boolqLunameasured1.00000.9900Valid0.899$0.000044
policy-021policyJevmeasured0.00000.0400Valid0.222$0.000021
policy-021policyLunameasured0.00000.0000Valid0.945$0.000046
boolq-dev-1114boolqJevmeasured1.00000.9900Valid0.224$0.000022
boolq-dev-1114boolqLunameasured1.00001.0000Valid1.075$0.000048
boolq-dev-3114boolqLunameasured1.00001.0000Valid1.467$0.000042
boolq-dev-3114boolqJevmeasured1.00000.9900Valid0.246$0.000019
probability-030probabilityJevmeasured0.45830.4100Valid0.219$0.000020
probability-030probabilityLunameasured0.45830.4167Valid1.065$0.000045
probability-007probabilityLunameasured0.72550.8214Valid1.260$0.000045
probability-007probabilityJevmeasured0.72550.4900Valid0.244$0.000020
probability-011probabilityJevmeasured0.54150.5900Valid0.227$0.000020
probability-011probabilityLunameasured0.54150.5840Valid1.412$0.000044
probability-025probabilityLunameasured1.00001.0000Valid1.061$0.000039
probability-025probabilityJevmeasured1.00000.9900Valid0.223$0.000018
boolq-dev-0767boolqLunameasured1.00000.5000Valid0.987$0.000048
boolq-dev-0767boolqJevmeasured1.00000.6200Valid0.261$0.000022
probability-057probabilityLunameasured0.55000.5500Valid0.988$0.000040
probability-057probabilityJevmeasured0.55000.5600Valid0.225$0.000018
boolq-dev-1447boolqJevmeasured1.00000.9700Valid0.226$0.000018
boolq-dev-1447boolqLunameasured1.00000.0000Valid1.260$0.000041
boolq-dev-2791boolqLunameasured0.00000.0000Valid0.908$0.000040
boolq-dev-2791boolqJevmeasured0.00000.0500Valid0.230$0.000018
probability-031probabilityLunameasured0.53570.5357Valid1.079$0.000041
probability-031probabilityJevmeasured0.53570.5200Valid0.346$0.000018
probability-033probabilityJevmeasured0.44210.5600Valid0.256$0.000018
probability-033probabilityLunameasured0.44210.4421Valid1.005$0.000041
boolq-dev-2493boolqJevmeasured1.00000.9600Valid0.268$0.000018
boolq-dev-2493boolqLunameasured1.00001.0000Valid1.180$0.000041
boolq-dev-1570boolqJevmeasured0.00000.0300Valid0.236$0.000022
boolq-dev-1570boolqLunameasured0.00000.0000Valid1.027$0.000048
policy-007policyLunameasured0.00000.0000Valid1.143$0.000046
policy-007policyJevmeasured0.00000.2300Valid0.231$0.000021
boolq-dev-1979boolqJevmeasured1.00000.9900Valid0.221$0.000020
boolq-dev-1979boolqLunameasured1.00001.0000Valid1.352$0.000045
probability-018probabilityJevmeasured0.73170.7700Valid0.224$0.000018
probability-018probabilityLunameasured0.73170.7585Valid1.445$0.000040
boolq-dev-0055boolqJevmeasured0.00000.3800Valid0.252$0.000044
boolq-dev-0055boolqLunameasured0.00000.1500Valid1.038$0.000103
boolq-dev-2504boolqJevmeasured0.00000.2300Valid0.242$0.000017
boolq-dev-2504boolqLunameasured0.00000.9800Valid1.050$0.000040
probability-039probabilityJevmeasured0.20000.2300Valid0.222$0.000018
probability-039probabilityLunameasured0.20000.2000Valid1.079$0.000040
probability-055probabilityJevmeasured0.70510.7200Valid0.262$0.000017
probability-055probabilityLunameasured0.70510.7051Valid1.041$0.000040
policy-013policyLunameasured1.00001.0000Valid1.017$0.000045
policy-013policyJevmeasured1.00000.9800Valid0.222$0.000020
policy-012policyJevmeasured1.00000.9800Valid0.234$0.000021
policy-012policyLunameasured1.00001.0000Valid1.270$0.000046
policy-025policyJevmeasured0.00000.0700Valid0.254$0.000021
policy-025policyLunameasured0.00000.0000Valid1.173$0.000046
policy-047policyJevmeasured0.00000.0400Valid0.254$0.000021
policy-047policyLunameasured0.00000.0000Valid0.982$0.000046
policy-041policyJevmeasured0.00000.2300Valid0.256$0.000021
policy-041policyLunameasured0.00000.0000Valid0.990$0.000046
boolq-dev-0448boolqJevmeasured1.00000.6100Valid0.221$0.000019
boolq-dev-0448boolqLunameasured1.00000.0000Valid0.978$0.000043
probability-042probabilityLunameasured0.00000.0000Valid1.276$0.000038
probability-042probabilityJevmeasured0.00000.0300Valid0.229$0.000018
policy-039policyJevmeasured1.00000.9800Valid0.226$0.000021
policy-039policyLunameasured1.00001.0000Valid1.592$0.000046
policy-036policyLunameasured0.00000.0000Valid1.003$0.000045
policy-036policyJevmeasured0.00000.0900Valid0.263$0.000020
boolq-dev-1625boolqJevmeasured1.00000.6600Valid0.222$0.000017
boolq-dev-1625boolqLunameasured1.00000.2500Valid1.260$0.000038
policy-040policyJevmeasured1.00000.9700Valid0.229$0.000021
policy-040policyLunameasured1.00001.0000Valid0.951$0.000046
policy-001policyJevmeasured1.00000.9800Valid0.279$0.000021
policy-001policyLunameasured1.00001.0000Valid1.065$0.000046
policy-028policyJevmeasured0.00000.2000Valid0.400$0.000021
policy-028policyLunameasured0.00000.0000Valid1.499$0.000046
probability-014probabilityJevmeasured0.95000.9300Valid0.253$0.000018
probability-014probabilityLunameasured0.95000.9500Valid1.437$0.000040
boolq-dev-3009boolqJevmeasured1.00000.9400Valid0.294$0.000019
boolq-dev-3009boolqLunameasured1.00001.0000Valid1.035$0.000041
boolq-dev-0341boolqJevmeasured0.00000.0300Valid0.233$0.000021
boolq-dev-0341boolqLunameasured0.00000.0000Valid1.478$0.000047
probability-020probabilityJevmeasured0.54590.3800Valid0.245$0.000020
probability-020probabilityLunameasured0.54590.4000Valid0.972$0.000044
probability-040probabilityLunameasured0.83170.5470Valid1.235$0.000044
probability-040probabilityJevmeasured0.83170.7200Valid0.239$0.000020
boolq-dev-3165boolqLunameasured1.00000.5000Valid1.256$0.000036
boolq-dev-3165boolqJevmeasured1.00000.6300Valid0.298$0.000016
probability-024probabilityLunameasured0.67590.5500Valid2.750$0.000044
probability-024probabilityJevmeasured0.67590.5000Valid0.248$0.000020
boolq-dev-1171boolqJevmeasured0.00000.2600Valid0.227$0.000017
boolq-dev-1171boolqLunameasured0.00000.0000Valid1.331$0.000038
probability-028probabilityJevmeasured0.82600.6800Valid0.226$0.000020
probability-028probabilityLunameasured0.82600.8372Valid0.999$0.000044
policy-027policyLunameasured0.00000.0000Valid1.063$0.000045
policy-027policyJevmeasured0.00000.0900Valid0.228$0.000020
policy-015policyJevmeasured1.00000.9800Valid0.246$0.000020
policy-015policyLunameasured1.00001.0000Valid1.016$0.000045
probability-017probabilityLunameasured0.51670.5167Valid1.451$0.000042
probability-017probabilityJevmeasured0.51670.5600Valid0.220$0.000018
boolq-dev-2552boolqJevmeasured1.00000.9900Valid0.259$0.000019
boolq-dev-2552boolqLunameasured1.00001.0000Valid1.140$0.000042
probability-052probabilityLunameasured0.48400.5247Valid1.052$0.000045
probability-052probabilityJevmeasured0.48400.5500Valid0.236$0.000020
probability-056probabilityJevmeasured0.75370.7700Valid0.221$0.000020
probability-056probabilityLunameasured0.75370.7283Valid1.880$0.000044
probability-032probabilityJevmeasured0.20000.2200Valid0.234$0.000018
probability-032probabilityLunameasured0.20000.2000Valid0.939$0.000040
boolq-dev-3125boolqJevmeasured1.00000.9700Valid0.216$0.000022
boolq-dev-3125boolqLunameasured1.00001.0000Valid1.294$0.000048
boolq-dev-2098boolqLunameasured1.00001.0000Valid1.501$0.000043
boolq-dev-2098boolqJevmeasured1.00000.9500Valid0.228$0.000019
policy-046policyLunameasured0.00000.0000Valid1.111$0.000045
policy-046policyJevmeasured0.00000.0700Valid0.264$0.000020
probability-048probabilityLunameasured0.07780.0778Valid0.945$0.000045
probability-048probabilityJevmeasured0.07780.1700Valid0.322$0.000020
policy-002policyJevmeasured1.00000.9800Valid0.228$0.000021
policy-002policyLunameasured1.00001.0000Valid1.478$0.000046
policy-056policyJevmeasured0.00000.0900Valid0.242$0.000021
policy-056policyLunameasured0.00000.0000Valid0.969$0.000046
policy-043policyLunameasured1.00001.0000Valid1.419$0.000045
policy-043policyJevmeasured1.00000.9800Valid0.334$0.000020
probability-005probabilityJevmeasured0.02500.0600Valid0.248$0.000018
probability-005probabilityLunameasured0.02500.0250Valid1.301$0.000040
probability-021probabilityJevmeasured0.20000.2200Valid0.254$0.000018
probability-021probabilityLunameasured0.20000.2000Valid1.372$0.000040
boolq-dev-1517boolqJevmeasured0.00000.8600Valid0.236$0.000019
boolq-dev-1517boolqLunameasured0.00001.0000Valid1.294$0.000043
probability-023probabilityJevmeasured0.19780.5800Valid0.224$0.000020
probability-023probabilityLunameasured0.19780.3600Valid1.014$0.000044
probability-022probabilityLunameasured0.69760.4446Valid1.017$0.000044
probability-022probabilityJevmeasured0.69760.6400Valid0.237$0.000020
boolq-dev-1859boolqJevmeasured0.00000.1700Valid0.237$0.000019
boolq-dev-1859boolqLunameasured0.00000.0500Valid1.245$0.000045
probability-027probabilityLunameasured0.23330.2333Valid1.102$0.000041
probability-027probabilityJevmeasured0.23330.3600Valid0.237$0.000018
policy-049policyJevmeasured1.00000.9900Valid0.262$0.000020
policy-049policyLunameasured1.00001.0000Valid0.953$0.000045
policy-034policyJevmeasured1.00000.9800Valid0.223$0.000021
policy-034policyLunameasured1.00001.0000Valid1.015$0.000046
boolq-dev-1335boolqLunameasured1.00001.0000Valid0.954$0.000057
boolq-dev-1335boolqJevmeasured1.00000.9200Valid0.388$0.000025
boolq-dev-1092boolqJevmeasured0.00000.0500Valid0.228$0.000020
boolq-dev-1092boolqLunameasured0.00000.0000Valid1.234$0.000046
policy-023policyJevmeasured1.00000.9800Valid0.255$0.000020
policy-023policyLunameasured1.00001.0000Valid1.302$0.000045
boolq-dev-2041boolqJevmeasured0.00000.0300Valid0.236$0.000021
boolq-dev-2041boolqLunameasured0.00000.0000Valid1.064$0.000047
probability-038probabilityLunameasured0.00000.0000Valid0.981$0.000037
probability-038probabilityJevmeasured0.00000.0500Valid0.227$0.000017
policy-057policyLunameasured0.00000.0000Valid1.272$0.000046
policy-057policyJevmeasured0.00000.0700Valid0.226$0.000021
probability-016probabilityLunameasured0.02370.0237Valid1.096$0.000041
probability-016probabilityJevmeasured0.02370.1500Valid0.234$0.000017
boolq-dev-0525boolqJevmeasured0.00000.0600Valid0.235$0.000016
boolq-dev-0525boolqLunameasured0.00000.0000Valid1.016$0.000036
boolq-dev-0980boolqLunameasured0.00000.0000Valid1.169$0.000038
boolq-dev-0980boolqJevmeasured0.00000.1100Valid0.215$0.000017
policy-060policyJevmeasured1.00000.9800Valid0.239$0.000021
policy-060policyLunameasured1.00001.0000Valid0.987$0.000046
policy-051policyLunameasured0.00000.0000Valid1.240$0.000045
policy-051policyJevmeasured0.00000.1100Valid0.318$0.000020
probability-060probabilityLunameasured0.49570.4467Valid1.581$0.000044
probability-060probabilityJevmeasured0.49570.5300Valid0.217$0.000020
probability-015probabilityLunameasured0.24520.2426Valid1.121$0.000040
probability-015probabilityJevmeasured0.24520.5900Valid0.288$0.000018
policy-052policyJevmeasured1.00000.9800Valid0.265$0.000021
policy-052policyLunameasured1.00001.0000Valid0.941$0.000046
policy-019policyLunameasured1.00001.0000Valid1.008$0.000046
policy-019policyJevmeasured1.00000.9800Valid0.246$0.000021
boolq-dev-2217boolqJevmeasured0.00000.0400Valid0.223$0.000019
boolq-dev-2217boolqLunameasured0.00000.0000Valid0.950$0.000042
boolq-dev-2011boolqLunameasured1.00001.0000Valid0.972$0.000054
boolq-dev-2011boolqJevmeasured1.00000.9900Valid0.223$0.000024
probability-010probabilityJevmeasured0.50000.5100Valid0.227$0.000018
probability-010probabilityLunameasured0.50000.5000Valid1.407$0.000040
boolq-dev-0877boolqJevmeasured1.00000.9300Valid0.274$0.000021
boolq-dev-0877boolqLunameasured1.00001.0000Valid1.018$0.000046
policy-032policyJevmeasured1.00000.9800Valid0.249$0.000021
policy-032policyLunameasured1.00001.0000Valid1.113$0.000046
policy-017policyJevmeasured1.00000.9800Valid0.268$0.000020
policy-017policyLunameasured1.00001.0000Valid2.314$0.000045
policy-008policyJevmeasured0.00000.0600Valid0.261$0.000021
policy-008policyLunameasured0.00000.0000Valid1.006$0.000046
boolq-dev-0065boolqJevmeasured1.00000.9600Valid0.381$0.000017
boolq-dev-0065boolqLunameasured1.00001.0000Valid1.104$0.000037
probability-045probabilityLunameasured0.50000.5000Valid0.998$0.000040
probability-045probabilityJevmeasured0.50000.5100Valid0.252$0.000018
probability-046probabilityLunameasured0.14130.0851Valid1.258$0.000044
probability-046probabilityJevmeasured0.14130.1500Valid0.248$0.000020
probability-001probabilityJevmeasured0.50000.5300Valid0.283$0.000018
probability-001probabilityLunameasured0.50000.5000Valid1.274$0.000040
probability-051probabilityJevmeasured0.50000.5200Valid0.237$0.000018
probability-051probabilityLunameasured0.50000.5000Valid1.156$0.000040
policy-011policyJevmeasured1.00000.9800Valid0.238$0.000021
policy-011policyLunameasured1.00001.0000Valid1.729$0.000046
policy-010policyLunameasured0.00000.0000Valid1.013$0.000045
policy-010policyJevmeasured0.00000.0700Valid0.319$0.000020
probability-043probabilityJevmeasured0.92590.8600Valid0.220$0.000020
probability-043probabilityLunameasured0.92590.9200Valid1.415$0.000044
policy-018policyLunameasured1.00001.0000Valid1.254$0.000045
policy-018policyJevmeasured1.00000.9800Valid0.263$0.000020
policy-009policyJevmeasured0.00000.0900Valid0.228$0.000020
policy-009policyLunameasured0.00000.0000Valid1.259$0.000045
boolq-dev-0659boolqLunameasured1.00001.0000Valid0.874$0.000037
boolq-dev-0659boolqJevmeasured1.00000.9900Valid0.231$0.000017
probability-034probabilityLunameasured0.98000.9800Valid1.044$0.000040
probability-034probabilityJevmeasured0.98000.9600Valid0.233$0.000018
boolq-dev-2951boolqLunameasured1.00001.0000Valid1.100$0.000058
boolq-dev-2951boolqJevmeasured1.00000.9700Valid0.259$0.000026
probability-053probabilityJevmeasured0.49910.6000Valid0.268$0.000018
probability-053probabilityLunameasured0.49910.5037Valid1.316$0.000039
probability-059probabilityJevmeasured0.22510.2000Valid0.311$0.000020
probability-059probabilityLunameasured0.22510.2489Valid1.389$0.000044
boolq-dev-1739boolqJevmeasured1.00000.8900Valid0.286$0.000020
boolq-dev-1739boolqLunameasured1.00001.0000Valid1.409$0.000044
policy-058policyLunameasured0.00000.0000Valid1.213$0.000046
policy-058policyJevmeasured0.00000.0800Valid0.221$0.000021
policy-053policyJevmeasured1.00000.9800Valid0.331$0.000020
policy-053policyLunameasured1.00001.0000Valid1.156$0.000045
probability-026probabilityJevmeasured0.87500.8900Valid0.219$0.000018
probability-026probabilityLunameasured0.87500.8750Valid1.007$0.000039
probability-009probabilityJevmeasured1.00000.9900Valid0.270$0.000018
probability-009probabilityLunameasured1.00001.0000Valid1.004$0.000039
boolq-dev-0308boolqLunameasured0.00000.0000Valid1.181$0.000043
boolq-dev-0308boolqJevmeasured0.00000.0500Valid0.255$0.000019
probability-050probabilityLunameasured0.51670.5167Valid1.158$0.000042
probability-050probabilityJevmeasured0.51670.5500Valid0.349$0.000018
boolq-dev-1296boolqJevmeasured0.00000.3900Valid0.372$0.000018
boolq-dev-1296boolqLunameasured0.00000.0000Valid1.168$0.000039
policy-044policyJevmeasured1.00000.9800Valid0.214$0.000021
policy-044policyLunameasured1.00001.0000Valid0.986$0.000046
policy-035policyLunameasured0.00000.0000Valid0.928$0.000045
policy-035policyJevmeasured0.00000.0600Valid0.235$0.000020
policy-031policyJevmeasured1.00000.9900Valid0.329$0.000020
policy-031policyLunameasured1.00001.0000Valid1.263$0.000045
probability-035probabilityJevmeasured0.20000.2200Valid0.291$0.000018
probability-035probabilityLunameasured0.20000.2000Valid1.281$0.000040
policy-026policyJevmeasured1.00000.9800Valid0.242$0.000021
policy-026policyLunameasured1.00001.0000Valid1.166$0.000046
probability-049probabilityLunameasured0.57720.5467Valid1.008$0.000044
probability-049probabilityJevmeasured0.57720.6300Valid0.238$0.000020
boolq-dev-2426boolqJevmeasured1.00000.9800Valid0.236$0.000022
boolq-dev-2426boolqLunameasured1.00001.0000Valid0.958$0.000050
probability-041probabilityJevmeasured0.00000.0300Valid0.257$0.000018
probability-041probabilityLunameasured0.00000.0000Valid1.106$0.000038
policy-045policyLunameasured0.00000.0000Valid1.190$0.000046
policy-045policyJevmeasured0.00000.0700Valid0.232$0.000021
policy-006policyLunameasured0.00000.0000Valid1.265$0.000046
policy-006policyJevmeasured0.00000.1100Valid0.327$0.000021
boolq-dev-3115boolqLunameasured1.00000.9800Valid1.143$0.000047
boolq-dev-3115boolqJevmeasured1.00000.9200Valid0.236$0.000021
policy-038policyJevmeasured0.00000.0300Valid0.246$0.000021
policy-038policyLunameasured0.00000.0000Valid1.389$0.000046
boolq-dev-1662boolqJevmeasured1.00000.9700Valid0.259$0.000021
boolq-dev-1662boolqLunameasured1.00001.0000Valid1.042$0.000046
policy-003policyLunameasured0.00000.0000Valid1.257$0.000046
policy-003policyJevmeasured0.00000.0500Valid0.227$0.000021
boolq-dev-2597boolqLunameasured1.00000.2500Valid0.970$0.000040
boolq-dev-2597boolqJevmeasured1.00000.9600Valid0.224$0.000018
probability-002probabilityJevmeasured0.78630.7700Valid0.235$0.000017
probability-002probabilityLunameasured0.78630.8194Valid1.234$0.000039
policy-059policyLunameasured1.00001.0000Valid0.900$0.000046
policy-059policyJevmeasured1.00000.9800Valid0.392$0.000021
probability-029probabilityJevmeasured0.88940.8700Valid0.232$0.000020
probability-029probabilityLunameasured0.88940.9250Valid1.338$0.000044
policy-050policyLunameasured1.00001.0000Valid1.204$0.000046
policy-050policyJevmeasured1.00000.9800Valid0.307$0.000021
boolq-dev-1381boolqLunameasured0.00000.0000Valid1.166$0.000037
boolq-dev-1381boolqJevmeasured0.00000.1500Valid0.317$0.000017
policy-048policyJevmeasured0.00000.1300Valid0.265$0.000021
policy-048policyLunameasured0.00000.0000Valid1.267$0.000046
boolq-dev-0210boolqLunameasured1.00001.0000Valid1.230$0.000041
boolq-dev-0210boolqJevmeasured1.00000.9600Valid0.352$0.000019
boolq-dev-2849boolqJevmeasured0.00000.4600Valid0.226$0.000018
boolq-dev-2849boolqLunameasured0.00000.0500Valid1.268$0.000042
probability-006probabilityLunameasured0.28000.5600Valid0.923$0.000039
probability-006probabilityJevmeasured0.28000.4100Valid0.272$0.000018
boolq-dev-2774boolqLunameasured0.00000.0500Valid1.185$0.000050
boolq-dev-2774boolqJevmeasured0.00000.3900Valid0.222$0.000023
probability-004probabilityJevmeasured0.20000.2200Valid0.257$0.000018
probability-004probabilityLunameasured0.20000.2000Valid1.562$0.000040
boolq-dev-0852boolqLunameasured0.00000.5000Valid0.978$0.000033
boolq-dev-0852boolqJevmeasured0.00000.1100Valid0.576$0.000014
boolq-dev-2868boolqJevmeasured0.00000.0700Valid0.226$0.000030
boolq-dev-2868boolqLunameasured0.00001.0000Valid1.192$0.000067
boolq-dev-3247boolqJevmeasured1.00000.9900Valid0.234$0.000018
boolq-dev-3247boolqLunameasured1.00001.0000Valid1.154$0.000039
boolq-dev-1603boolqJevmeasured0.00000.0400Valid0.267$0.000020
boolq-dev-1603boolqLunameasured0.00000.0000Valid1.047$0.000045
probability-058probabilityJevmeasured0.66910.6900Valid0.241$0.000020
probability-058probabilityLunameasured0.66910.7989Valid1.143$0.000044
policy-014policyLunameasured0.00000.0000Valid1.926$0.000045
policy-014policyJevmeasured0.00000.0900Valid0.244$0.000020
policy-020policyJevmeasured0.00000.0900Valid0.234$0.000021
policy-020policyLunameasured0.00000.0000Valid1.210$0.000046
boolq-dev-2116boolqLunameasured1.00001.0000Valid1.306$0.000043
boolq-dev-2116boolqJevmeasured1.00000.9900Valid0.221$0.000019
policy-033policyJevmeasured0.00000.1300Valid0.244$0.000021
policy-033policyLunameasured0.00000.0000Valid1.214$0.000046
policy-029policyLunameasured0.00000.0000Valid0.898$0.000046
policy-029policyJevmeasured0.00000.0900Valid0.220$0.000021
policy-055policyJevmeasured1.00000.9700Valid0.221$0.000021
policy-055policyLunameasured1.00001.0000Valid1.244$0.000046
policy-005policyJevmeasured1.00000.9800Valid0.234$0.000020
policy-005policyLunameasured1.00001.0000Valid1.199$0.000045
policy-030policyJevmeasured0.00000.0700Valid0.238$0.000021
policy-030policyLunameasured0.00000.0000Valid1.048$0.000046
probability-012probabilityJevmeasured0.88900.8600Valid0.224$0.000018
probability-012probabilityLunameasured0.88900.8654Valid1.333$0.000040
policy-022policyLunameasured1.00001.0000Valid1.073$0.000046
policy-022policyJevmeasured1.00000.9800Valid0.219$0.000021
probability-013probabilityLunameasured0.95000.9500Valid1.142$0.000040
probability-013probabilityJevmeasured0.95000.9400Valid0.234$0.000018
boolq-dev-3134boolqLunameasured1.00001.0000Valid1.563$0.000036
boolq-dev-3134boolqJevmeasured1.00000.9800Valid0.228$0.000016
boolq-dev-1953boolqLunameasured0.00000.0000Valid1.470$0.000038
boolq-dev-1953boolqJevmeasured0.00000.0200Valid0.279$0.000017
policy-054policyLunameasured1.00001.0000Valid1.027$0.000046
policy-054policyJevmeasured1.00000.9800Valid0.229$0.000021
boolq-dev-2818boolqJevmeasured1.00000.8800Valid0.264$0.000019
boolq-dev-2818boolqLunameasured1.00000.5000Valid1.021$0.000044
policy-004policyLunameasured0.00000.0000Valid1.642$0.000045
policy-004policyJevmeasured0.00000.0700Valid0.230$0.000020
boolq-dev-0383boolqJevmeasured1.00000.9900Valid0.224$0.000017
boolq-dev-0383boolqLunameasured1.00001.0000Valid1.119$0.000037
probability-037probabilityLunameasured0.03330.0333Valid1.499$0.000043
probability-037probabilityJevmeasured0.03330.0600Valid0.234$0.000018
probability-036probabilityJevmeasured0.10480.2800Valid0.283$0.000017
probability-036probabilityLunameasured0.10480.1000Valid1.063$0.000038
probability-003probabilityLunameasured0.00000.0000Valid1.301$0.000039
probability-003probabilityJevmeasured0.00000.0100Valid0.254$0.000018
boolq-dev-2665boolqLunameasured0.00000.9700Valid0.960$0.000046
boolq-dev-2665boolqJevmeasured0.00000.2200Valid0.338$0.000020
boolq-dev-1842boolqLunameasured1.00000.0000Valid1.265$0.000041
boolq-dev-1842boolqJevmeasured1.00000.0300Valid0.352$0.000018
probability-008probabilityLunameasured0.96920.9677Valid1.448$0.000040
probability-008probabilityJevmeasured0.96920.9600Valid0.220$0.000018
probability-047probabilityLunameasured0.77080.8000Valid1.075$0.000044
probability-047probabilityJevmeasured0.77080.7300Valid0.222$0.000020
policy-024policyJevmeasured0.00000.0900Valid0.219$0.000020
policy-024policyLunameasured0.00000.0000Valid1.049$0.000045
boolq-dev-0468boolqJevmeasured1.00000.9900Valid0.225$0.000018
boolq-dev-0468boolqLunameasured1.00000.9900Valid1.136$0.000042
boolq-dev-0587boolqJevmeasured1.00000.1500Valid0.234$0.000019
boolq-dev-0587boolqLunameasured1.00000.0000Valid1.318$0.000041
boolq-dev-2630boolqLunameasured1.00001.0000Valid0.992$0.000039
boolq-dev-2630boolqJevmeasured1.00000.9800Valid0.240$0.000018
probability-019probabilityJevmeasured0.93330.9100Valid0.267$0.000018
probability-019probabilityLunameasured0.93330.9333Valid2.471$0.000041
policy-037policyJevmeasured1.00000.9800Valid0.270$0.000021
policy-037policyLunameasured1.00001.0000Valid1.001$0.000046
probability-054probabilityJevmeasured0.02380.1500Valid0.246$0.000017
probability-054probabilityLunameasured0.02380.0238Valid1.023$0.000040
policy-042policyLunameasured1.00001.0000Valid1.174$0.000046
policy-042policyJevmeasured1.00000.6000Valid0.244$0.000021
Metric definitions and statistical method
success
valid=true with a finite probability and valid target in [0,1]; label targets must be 0 or 1
accuracy
Fraction correct among valid responses, predicting yes when p_yes >= 0.5
accuracy_all_requests
Correct valid responses divided by every measured request; invalid requests count as failures
brier
Mean (p_yes - observed_label)^2; lower is better
log_loss
Mean binary negative log likelihood in natural-log units; p clipped to [1e-15, 1-1e-15]
mae
Mean absolute difference between reported p_yes and exact reference probability
rmse
Square root of mean squared error against exact reference probabilities
excess_brier
Mean (p_yes - true_probability)^2: expected Brier above the optimal probability forecast
expected_brier
Mean [(p_yes - true_probability)^2 + true_probability*(1-true_probability)]
latency
End-to-end request time through receipt of a complete response; main table includes valid responses only
cost
Recorded billed USD across all requests in the stated phase, including failed/invalid requests when billed
cost_per_1000
Measured total billed cost / measured request count * 1000; withheld if any bill is missing
cost_per_correct
All measured billed cost / number of correct valid label responses; withheld for missing bills or zero correct
bootstrap
Percentile 95% paired bootstrap by case identity; quality repeats averaged within case; latency repeats use median; no CI for fewer than 2 pairs
Recorded run metadata
{
  "models": {
    "Jev": "typesafe/jev-1.13",
    "Luna": "openai/gpt-6-luna"
  },
  "created_at": "2026-10-02T08:42:23.244948+00:00",
  "seed": 20261002,
  "cases": 180,
  "calls_planned": 360,
  "budget_usd": 0.999912658,
  "luna_reasoning": "none",
  "luna_provider": "OpenAI",
  "protocol": "One non-streaming request per model per case; sequential paired random order; one shared persistent HTTPS connection; no automatic retries; two warmups per model excluded from measured results.",
  "cases_sha256": "60e6db60c72869936548c3c4620aa3b070cf12e516f3be42a61a306d0f592660",
  "runner_sha256": "3a05cb20816fbc5b7c30a065429a7365afa9583f273d683bf69784f4af3e9ad7",
  "status": "complete",
  "charged_usd": 0.011550069999999994,
  "calls_completed": 364,
  "finished_at": "2026-10-02T08:46:47.575786+00:00"
}