gpt-oss-safeguard:20b

Upstream reference: gpt-oss-safeguard

Publisher: OpenAI and ROOST

The official guide describes gpt-oss-safeguard as a policy-following text-classification model derived from gpt-oss, with written policies supplied by the application.

Publisher-described context, separate from Sigma5C observations and recorded catalog statistics.

Published upstream specifications

Model role

Policy-based text classification Source

Documented output design

Reasoning and formatted classification channels using Harmony Source

Publisher-described uses

• Classifying text against a supplied content policy; policy testing and review queues. Source

Reference limitations

• Its output depends on the supplied policy and thresholds; a classification is not an independent legal or safety determination. Source

A recorded tag and an upstream card association do not verify the deployed checkpoint, quantization, configured context or current hosted availability.

Publisher sources

• gpt-oss-safeguard official documentation — OpenAI and ROOST. Checked 2026-11-02.

Links

Recorded references do not establish downloadable weights, licensing or current availability.

• OpenAI — official site
• Ollama
• Provider page
• Hugging Face

Where gpt-oss-safeguard:20b comes from

Named in 1 observation page

• 14B-32B local models — all 47 · V1