LuminaBench
Reasoning benchmark / Conceptual Reasoning

DTBENCH

Lumina VerifiedLumina checked the published sources, versions, model labels, score extraction and conversion. External evaluations were run by their publishers; Lumina has not independently rerun or reproduced them.

Decision accuracy

Version CRI-public-components-20261106Verified 6 November 2026268 published resultsOfficial source

DTBENCH

LuminaBench
Show
20 of 212 models · best variant per model
03006009001,2001,115Claude Opus 5.5Max1,113Claude Fable 5Max1,110Claude Opus 5xHigh1,109Gemini 4 ArgonLow1,109Claude Fable 5.1xHigh1,108Claude Sonnet 5.5Max1,108GPT-6 AstraxHigh1,108GPT-6 Sol1,108Grok 4.61,107Gemini 3.1 ProMedium1,106Gemini 3.7 FlashHigh1,105GPT-6.1 Sol1,105Grok 4.51,105Muse Spark 1.31,102Grok 4.71,102GPT-5.6 Sol Pro1,102GPT-5.5 Pro1,102GPT-5.51,101GPT-5.6 SolxHigh1,101Gemini 3.8 FlashHigh

Lumina Score vs Cost per Task

Original average USD per LMCA/DTBench response as documented in the source data README; absent costs stay unavailable. No time or token measurements inferred.

LuminaBench
15 of 156 families
Lower resource use and higher score (green); higher resource use and lower score (red). Guides split at the visible selection’s medians.$0.01$0.03$0.1$0.38009001,0001,100Cost per task · USD (log scale)Lumina ScoreClaude Opus 5.5Claude Fable 5Claude Opus 5Claude Sonnet 5.5Gemini 4 ArgonGemini 3.7 Flash

Original average USD per LMCA/DTBench response as documented in the source data README; absent costs stay unavailable. No time or token measurements inferred. Missing metrics are omitted from their graph; quality scores remain available.

Results table

Method & coverage

Original task component from public cri-scores.csv. Preserve native efforts and assisted/partial qualifications. Parent CRI aggregate is not imported. DTBench original decision accuracy, not its rescaled CRI component.

Original average USD per LMCA/DTBench response as documented in the source data README; absent costs stay unavailable. No time or token measurements inferred.

How the Lumina conversion works

1,000 + 75 × ((original score − 72.45417138364783) / 17.25220649988741). The result is rounded once. Scores may tie.

Frozen initial observed text-only cohort; one mean per exact publisher release/API, balancing repeated settings/routes within that release. Original score and protocol remain separate. No inferred variants or future refitting.

Calibration
dtbench-z75-v1 · experimental
Reference mean
72.45417138364783
Population SD
17.25220649988741
Coverage
Standalone benchmark.
Identity, exclusions & attribution

DTBench rescaled uncertainty is not attached to raw decision accuracy; ACCoRD rescaled uncertainty is not attached to normalized constraint violation.

Text conceptual comparisons, decision problems and logical consistency constraints; no visual or audio inputs.

Standalone text benchmark; no new combined index. Source identity/effort/protocol qualifications remain in Data and CSV.

Official source · Publication terms / evidence

Publisher terms. Scores and resource figures are attributed to the original publisher.