TEMPORAL COMPARISON

Same Test. Latest view + history.

Compare 2–4 canonical models against one public Test Case, then inspect each model’s sealed Run history without turning the Archive into a ranking table.

01 / Public TestLatest comparison and temporal lanes use this exact public Test Case.Comparability policyVersioned eligibility rules, not a model score. Earlier policy versions remain selectable for re-evaluating historical Runs.

02 / Models

DeepSeek Flashdeepseek · deepseek-flash

1 / 4 selected · minimum 2

DeepSeek FlashNo winner, score aggregation, or ordering is computed.

SIDE BY SIDE

Select models and run comparison

Comparability is evaluated under a versioned policy. Calibration canaries are service-health evidence, not leaderboard scores or a degradation verdict.

Comparison intentionally uses raw Archive records. It does not infer an overall model winner from heterogeneous Tests.

SAME-TEST HISTORY

Temporal lanes appear after comparison

Every lane is ordered by capture time. Result changes are displayed as observations, not as an inferred trend, winner, or cause.

Select one Test and 2–4 models to inspect both the latest sealed record and the sequence of earlier sealed Runs for the same exact Test.