Model family

gpt-oss-safeguard

gpt-oss-safeguard is a model family from OpenAI1 2, with 2 assessed releases in this catalog. Licenses and availability are recorded for each release, not for the family; its releases use the Apache License 2.0.

gpt-oss-safeguard is a pair of open-weight, text-only reasoning models from OpenAI, gpt-oss-safeguard-120b and gpt-oss-safeguard-20b, post-trained from the corresponding gpt-oss models to classify content against a policy that the user supplies. OpenAI's technical report is dated October 29, 2025. OpenAI lists the models as a model partner of the Robust Open Online Safety Tools (ROOST) Model Community, a group of safety practitioners working with open models.1 2 3 4Fact reviewed Oct 1, 2026

Repository: gpt-oss-safeguard repository (external site: github.com)
Model hub: gpt-oss-safeguard-120b on Hugging Face (external site: huggingface.co)
Model hub: gpt-oss-safeguard-20b on Hugging Face (external site: huggingface.co)
Paper: Technical report (PDF) (external site: cdn.openai.com)
Documentation: gpt-oss-safeguard user guide (external site: developers.openai.com)

Last reviewedOct 1, 2026Entry updated Oct 1, 2026Documented release Oct 2025

Releases assessed

Each release is assessed on its own. This family has no family-wide openness label.

Releases in the gpt-oss-safeguard family
Release	Weights	Tier (USASI rubric v0.2)	License	Released
gpt-oss-safeguard-120b	Weights: Public	Model-disclosure tier (USASI rubric v0.2): Open-weight	Apache-2.0, Apache-2.0	Oct 2025
gpt-oss-safeguard-20b	Weights: Public	Model-disclosure tier (USASI rubric v0.2): Open-weight	Apache-2.0, Apache-2.0	Oct 2025

What it is useful for

The model cards list safety uses such as filtering LLM inputs and outputs, online content labeling, and offline labeling for trust-and-safety work. OpenAI recommends the base gpt-oss models for other applications, including end-user chat.3 1Fact reviewed Oct 1, 2026

Organization context

OpenAI

Provenance and derivatives

Both models are fine-tunes of OpenAI's gpt-oss models. The technical report says they were trained without any additional biological or cybersecurity data.1 3 4

Derived from: gpt-oss — Base model family (gpt-oss-120b and gpt-oss-20b).

U.S. eligibility

Project eligibility rests on documented governing or maintaining entities, not on contributors.

Eligible · basis: U.S. headquarters

Developed and published by OpenAI. Its operating company, OpenAI Group PBC, is described as a Delaware public benefit corporation at 1455 3rd Street, San Francisco, California, in a February 2026 agreement filed with the SEC.1 7

Assessed Oct 1, 2026

Sources

1.
Technical Report: Performance and baseline evaluations of gpt-oss-safeguard-120b and gpt-oss-safeguard-20b (external site: cdn.openai.com)
OpenAI · Paper · published Oct 29, 2025 · accessed Oct 1, 2026 · evidence reviewed Oct 1, 2026
2.
openai/gpt-oss-safeguard README (external site: raw.githubusercontent.com)
OpenAI · Repository · accessed Oct 1, 2026 · evidence reviewed Oct 1, 2026
3.
openai/gpt-oss-safeguard-120b model card (external site: huggingface.co)
OpenAI (Hugging Face) · Model card · accessed Oct 1, 2026 · evidence reviewed Oct 1, 2026
4.
openai/gpt-oss-safeguard-20b model card (external site: huggingface.co)
OpenAI (Hugging Face) · Model card · accessed Oct 1, 2026 · evidence reviewed Oct 1, 2026
5.
Hugging Face model metadata for openai/gpt-oss-safeguard-120b (external site: huggingface.co)
Hugging Face · Repository · accessed Oct 1, 2026 · evidence reviewed Oct 1, 2026
6.
Hugging Face model metadata for openai/gpt-oss-safeguard-20b (external site: huggingface.co)
Hugging Face · Repository · accessed Oct 1, 2026 · evidence reviewed Oct 1, 2026
7.
Exhibit 10.1: Equity commitment letter agreement between OpenAI Group PBC and Amazon (external site: sec.gov)
U.S. Securities and Exchange Commission (Amazon.com, Inc. filing) · Filing · published Feb 27, 2026 · accessed Oct 1, 2026 · evidence reviewed Oct 1, 2026

Email a correction Report a correction Edit this entry

This listing is not an endorsement, a safety assessment, or a federal approval.

Cite this page

APA style

USASI. (2026, October 1). gpt-oss-safeguard. United States of America Superintelligence. https://unitedstatesofamericasuperintelligence.com/open/gpt-oss-safeguard/

BibTeX

@misc{usasi_gpt_oss_safeguard,
  author = {{USASI}},
  title = {gpt-oss-safeguard},
  year = {2026},
  month = oct,
  howpublished = {\url{https://unitedstatesofamericasuperintelligence.com/open/gpt-oss-safeguard/}},
  note = {United States of America Superintelligence. Last updated 2026-10-01}
}

The date is when this page was last updated. Add the date you read it if your style needs one.

For maintainers: a “Listed in USASI” badge

Links to this page. It says the entry is listed, not endorsed or rated. About the badge

Markdown (for a README)

[![Listed in the USASI catalog](https://unitedstatesofamericasuperintelligence.com/badges/listed.svg)](https://unitedstatesofamericasuperintelligence.com/open/gpt-oss-safeguard/)

HTML

<a href="https://unitedstatesofamericasuperintelligence.com/open/gpt-oss-safeguard/"><img src="https://unitedstatesofamericasuperintelligence.com/badges/listed.svg" alt="Listed in the USASI catalog" height="20"></a>