Skip to content
RankShark

gpt-oss-safeguard-20b

gpt-oss-safeguard-20b is a safety reasoning model from OpenAI built upon gpt-oss-20b.

ModelOpen sourceby OpenAIsince 2025
26.9

SharkScore

#359 in Models

Why it scores 26.9

gpt-oss-safeguard-20b has a SharkScore of 26.9/100, ranked #359 of 608 in Models (updated 2026-11-19).

Each part compares gpt-oss-safeguard-20b with other listings in Models. Evidence covers 87% of the score's weight.

Adoption
8 · 30%
User reviews
no data · 25%
Maintenance
4 · 20%
Trust & transparency
80 · 15%
Momentum
no data · 10%

Paying never changes this score. How ranking works · method v2026-11-18.2

RankShark review of gpt-oss-safeguard-20b

Automated evaluation by RankShark · 18 Nov 2026 · method v1.0 · How we review

We were not paid for this review.

As of 2026-11-18, gpt-oss-safeguard-20b is an open-weight safety reasoning model; the OpenRouter listing measures 131,072 tokens of context and charges per token. [E1, E2, MEASURED]

OpenAI describes gpt-oss-safeguard-20b as a model for classifying text against user-provided safety policies and other safety tasks. The model card says it has 21B parameters and is built on gpt-oss. [E1]

Best for

  • Teams building text classification or filtering around their own safety policies. · model card
  • Safety practitioners who want to run model weights on GPUs with 16GB of VRAM. · model card

Not for

  • General-purpose applications outside safety use cases; the model card recommends gpt-oss models for other applications. · model card
  • Systems that cannot use the Harmony response format; the model card says the model should only be used with that format. · model card

Key features

  • Safety classification. The vendor says the model can classify text using policies supplied by the user. · model card
  • Reasoning output. The vendor says the model provides access to its reasoning process for debugging policy decisions. · model card
  • Configurable reasoning effort. The vendor says reasoning effort can be set to low, medium, or high. · model card
  • Hardware requirement. The vendor says the 21B-parameter model fits on GPUs with 16GB of VRAM. · model card
  • Context window. The OpenRouter listing gives a 131,072-token context window; RankShark measured the same value on 2026-11-18. · OpenRouter

Pros

  • The model card lists an Apache 2.0 license. · model card
  • The model supports written user policies and multiple reasoning-effort settings, according to the vendor. · model card

Cons

  • The model card says it should only be used with the Harmony response format. · model card
  • The vendor positions the model for safety use cases and recommends gpt-oss models for other applications. · model card
  • No free plan or free trial stated on the pages we checked.
  • Whether customer data is used for model training is not stated on the pages we checked.

Pricing

PlanPriceLimits
OpenRouter token pricing$0.075 per million input tokens; $0.30 per million output tokensPer-token pricing; no monthly plan or monthly limit is stated.

Checked 18 Nov 2026

Free plan
Not stated (checked 18 Nov 2026)
Free trial
Not stated (checked 18 Nov 2026)
Trains on customer data
Not stated (checked 18 Nov 2026)
Data retention
Not stated (checked 18 Nov 2026)
Data regions
Not stated (checked 18 Nov 2026)
Data processing agreement
Not stated (checked 18 Nov 2026)
AI models used
gpt-oss-safeguard-20b, a 21B-parameter model built on gpt-oss-20b. · model card, OpenRouter
API
The OpenRouter listing includes chat completions, completions, messages, and responses endpoints. · OpenRouter
SDKs
Not stated (checked 18 Nov 2026)
Webhooks
Not stated (checked 18 Nov 2026)
Platforms
The model is listed on Hugging Face; its model card names Transformers and includes a vLLM tag. · model card
Security claims
No security certification stated (checked 18 Nov 2026)

What we measured

Context window (OpenRouter listing) · now
131,072 tokens
Price per million tokens (OpenRouter listing) · now
$0.075 in, $0.3 out
Package downloads · 30 days
29,259
How we tested · sources
  • Hourly uptime and response-time checks of the product's URL.
  • Repository, release and download data from GitHub, npm, PyPI and Hugging Face where they exist.
  • We did not sign up, pay, or use the product as a customer.
  • 18 Nov 2026: Plans or prices changed on the pricing page.
  • 18 Nov 2026: Plans or prices changed on the pricing page.

Something wrong? Report an error with a public link that shows it; corrections re-run the review.

Top alternatives to gpt-oss-safeguard-20b

See all
  • 67.1

    Qwen3.8 27B is an open-weight dense vision-language model from Qwen.

    ModelOpen source
  • 65.5

    Qwen3.6-35B-A3B is an open-weight multimodal model from Alibaba Cloud with 35 billion total parameters and 3 billion active parameters per t

    ModelOpen source
  • 64.7

    Qwen3.8 Flash is a multimodal reasoning model from Alibaba.

    ModelOpen source
  • 64.3

    Qwen3.6 27B is a dense 27-billion-parameter language model from the Qwen Team at Alibaba, released in April 2026.

    ModelOpen source