Models / openai / GPT-OSS Safeguard 20B

GPT-OSS Safeguard 20B

JSON →
openaillm

A 20-billion-parameter safety classifier from OpenAI designed to detect harmful content and enforce usage policies in open-source deployments.

textjson-modestreamingopen-weightsfunction-callingtool-usereasoning
Specs
Context & limits
131K
context window
66K
max output
Lifecycle
Dates
released2025-03-01
knowledge cutoff2024-06-01
Resources