deepseek-v4.1-flash-1017@deepinfra
Measured against its own past on the daily and weekly batteries. Each battery has its own history and assessment.
| provider channel | openai_compat |
|---|---|
| identity | pinned snapshot · a dated id that promises stability; a change here means the snapshot itself changed |
| daily battery | 20 readings (2026-10-21 to 2026-11-09) Latest assessment: Assessed (13/13 metrics eligible) |
| weekly battery | 3 readings (2026-10-25 to 2026-11-08) Latest assessment: Baseline accruing (0/13 metrics eligible) |
Published findings
Daily: No findings in completed eligible tests on the latest reading.
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.
First reading reproduces the deepseek-chat regression · 2026-10-21
This is the dated DeepSeek-V4.1-Flash-1017 snapshot served by DeepInfra, added the day after DeepSeek's own API updates page dated the 1017 upgrade (2026-10-17). On its first reading it failed the same two instruction-following probes, in the same way, as the deepseek-chat alias has since 2026-10-17: an exclusion constraint ignored and a word-count floor undershot. Different host, different key, same weights. Recorded as a reproduction on an independent channel; the instrument does not establish why the build behaves this way.
Explore
Open the daily dossier · Open the weekly dossier. Each dossier shows this model and battery, with the full dimension history.