Can enterprise Gemini APIs disable safety filters? Devs probe HarmBlockThreshold settings
moschles · reddit · 2026-10-01
An enterprise team on Reddit asks whether Gemini models invoked via enterprise APIs can be configured with HarmBlockThreshold = BLOCKNONE and finer control like temperature, to minimize censorship for internal agent workloads.
The post lists rumored harm category enums — HARMCATEGORYHARASSMENT, HATESPEECH, SEXUALLYEXPLICIT, DANGEROUSCONTENT — and notes a finishreason enum value called RECITATION, where the model blocks requests over copyrighted material. The author asks whether more categories exist and how well these settings actually work.
More from Models
- Leak Claims OpenAI Killed Its Most Capable Model After Safety Tests — YvesMulkers · 2026-10-01
- Sebastian Raschka: Jev is a text classifier, but not 'just' a classifier — neal_lathia · 2026-10-01
- "There'll always be a new benchmark to conquer," VraserX replies to benchmark fatigue — VraserX · 2026-10-01
- Beam Launches Playground to Try Open-Source Decision Models as Cloud APIs — velobro · 2026-10-01
- Why Do Benchmark Scores Rise Every Release? Reddit Debates Closed Evals — doomadah · 2026-10-01
- Marathoner: a 9B model that codes for 10+ hours, lifting SWE-bench Verified to 77.5% — mark_k · 2026-10-01