gemini-3.8-flash
Measured against its own past on the daily and weekly batteries. Each battery has its own history and assessment.
| provider channel | |
|---|---|
| identity | alias · a floating name the provider may re-point; a change here means what is served under this name changed |
| daily battery | 41 readings (2026-09-30 to 2026-11-09) Latest assessment: Assessment limited (9/10 metrics eligible) |
| weekly battery | 6 readings (2026-10-04 to 2026-11-08) Latest assessment: Baseline accruing (0/9 metrics eligible) |
Published findings
Daily: Only some metrics could be assessed; the rest lacked baseline or coverage.
Weekly: No assessment completes until enough baseline readings accumulate.
Editorial notes
An annotation attaches a dated external record (a provider changelog entry, a documented deprecation, a known configuration artefact) to a reading the instrument already produced. It changes no reading and establishes no cause; the instrument detects, attribution stays a human judgment.
Configuration artefact in the first segment · 2026-10-12
Readings from 2026-10-12 to 2026-10-21 were taken with a thinking setting this model ignores (thinking_budget 0; Gemini 3.x models take thinking_level, default medium, with no off setting per Google's thinking documentation). The model reasoned at its default level and consumed the small output budgets of the short-answer probes: 9 to 14 calls per reading returned no visible text (recorded as truncation call errors) and other short answers were cut off, which depressed the capability and instruction-following pass rates. Roster v0.31 (2026-10-21) set thinking_level low; the series restarts there. The first segment reflects the instrument's configuration, not the model's behaviour.
Explore
Open the daily dossier · Open the weekly dossier. Each dossier shows this model and battery, with the full dimension history.