AIQ v1 fixed-fixture evaluation

A score is only useful
when you can inspect it.

AIQ Wiki pairs each fixed-fixture result with task sensitivity, coverage, failures, history, and a route to all 72 task outcomes.

AIQ
Publication statusNo published score yetThe 17 configurations remain visible until a complete run is verified.
Matrix176 Sol · 6 Terra · 5 Luna
Task observations0across scored entries
Mean coverageUnknownnot measured
Radar nodes3identity + provenance
Model × reasoning matrix

Public index

Official results require non-synthetic evidence for all 72 committed fixtures. Bounds are task-resampling sensitivity intervals, not general capability confidence intervals. The table order is descriptive and does not establish a winner.

No published evidence

The live public read is available, but it has no evidence to display.

Descriptive order by fixed-fixture point estimate, high to low. This order does not identify a statistical winner.
Model / reasoningScoreTask sensitivitySamplesCoverageFailed / missingScoringCompletenessEvidenceRun
Lunagpt-5.6-luna · highUnpublishedNo evidence
Lunagpt-5.6-luna · lowUnpublishedNo evidence
Lunagpt-5.6-luna · maxUnpublishedNo evidence
Lunagpt-5.6-luna · mediumUnpublishedNo evidence
Lunagpt-5.6-luna · xhighUnpublishedNo evidence
Solgpt-5.6-sol · highUnpublishedNo evidence
Solgpt-5.6-sol · lowUnpublishedNo evidence
Solgpt-5.6-sol · maxUnpublishedNo evidence
Solgpt-5.6-sol · mediumUnpublishedNo evidence
Solgpt-5.6-sol · ultraUnpublishedNo evidence
Solgpt-5.6-sol · xhighUnpublishedNo evidence
Terragpt-5.6-terra · highUnpublishedNo evidence
Terragpt-5.6-terra · lowUnpublishedNo evidence
Terragpt-5.6-terra · maxUnpublishedNo evidence
Terragpt-5.6-terra · mediumUnpublishedNo evidence
Terragpt-5.6-terra · ultraUnpublishedNo evidence
Terragpt-5.6-terra · xhighUnpublishedNo evidence
Look beyond the order

Compare like with like.

Select exact model and reasoning combinations. Inspect descriptive differences, intervals, and compatibility.

Open comparison studio
History is evidence

Keep the whole timeline.

Switch among day, week, month, and all-history views without deleting the past.

Explore score history