deepseek-reasoner@deepseek
Measured against its own past on the daily and weekly batteries. Each battery has its own history and assessment.
| provider channel | openai_compat |
|---|---|
| identity | alias · a floating name the provider may re-point; a change here means what is served under this name changed |
| daily battery | 79 readings (2026-08-23 to 2026-11-09) Latest assessment: Assessed (13/13 metrics eligible) |
| weekly battery | 13 readings (2026-08-27 to 2026-11-08) Latest assessment: Assessed (13/13 metrics eligible) |
Published findings
Thinking-token use rose on the sycophancy probe set Movement
Thinking-token use rose on the verbosity probe set Movement
Tool call pass rate fell Movement
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.
Provider record · 2026-10-17
DeepSeek's API updates page, 2026-10-17: DeepSeek-V4.1-Flash upgraded to DeepSeek-V4.1-Flash-1017, with shorter outputs. From the 2026-10-17 reading this alias's mean thinking tokens on the verbosity probes fell from about 1,000 to about 200 and on the sycophancy probes from about 1,100 to about 600. Dated record alongside the readings; cause not established by the instrument.
Explore
Open the daily dossier · Open the weekly dossier. Each dossier shows this model and battery, with the full dimension history.