Skip to main content

ModelRegistry

Explore text, image, video and speech models

Find the right AI by model, modality and provider.

Browse models

Choose with facts

Turn the model directory into an actionable evaluation workflow.

Clear capabilities

Review modalities, protocols, limits and status.

Transparent coverage

See which providers expose each model.

Easy comparison

Compare candidates using consistent dimensions.

Start exploring the directory

Begin with a model, provider or organization.

Open model directory

Model spotlight

gpt-oss-safeguard-20b

gpt-oss-safeguard-20b is a safety reasoning model from OpenAI built upon gpt-oss-20b. This open-weight, 21B-parameter Mixture-of-Experts (MoE) model offers lower latency for safety tasks like content classification, LLM filtering, and trust...

Compare models

Core capabilities

A compact view of the model contract before integration.

Input and output

See supported modalities and generation directions.

Limits and context

Check context, output limits and release details.

Provider coverage

Compare the providers and protocols that expose this version.

gpt-oss-safeguard-20b FAQ

Start building with gpt-oss-safeguard-20b

Move from model evaluation to a production-ready workflow.

Explore providers

gpt-oss-safeguard-20b

openai/gpt-oss-safeguard-20b

gpt-oss-safeguard-20b is a safety reasoning model from OpenAI built upon gpt-oss-20b. This open-weight, 21B-parameter Mixture-of-Experts (MoE) model offers lower latency for safety tasks like content classification, LLM filtering, and trust...

Model family

GPT

-

Focus areas

ReasoningGenerative AI

Core specifications

Released
2025-10-29
Context window
131K
Max output
66K
Model parameters
22B
Weights
Open

Modalities

Input
Text
Output
Text

Identity

Version
gpt-oss-safeguard-20b
License
Apache-2.0

Resources

Capabilities

Open weightReasoningStructuredTemperatureTool call

Capabilities and fit

Model summary

gpt-oss-safeguard-20b is a safety reasoning model from OpenAI built upon gpt-oss-20b. This open-weight, 21B-parameter Mixture-of-Experts (MoE) model offers lower latency for safety tasks like content classification, LLM filtering, and trust...

Text
Open weightReasoningStructuredTemperatureTool call
Output
Text
Capabilities
5
Catalog updated
2026-08-06T01:13:30.619Z
Status

Provider coverage

ProviderInputOutputPricing
OpenRouter
gpt-oss-safeguard-20b
Pricing unavailable
NanoGPT
GPT OSS Safeguard 20B
Pricing unavailable
Groq
Safety GPT OSS 20B
Pricing unavailable
Kilo Gateway
OpenAI: gpt-oss-safeguard-20b
Pricing unavailable
Vercel AI Gateway
gpt-oss-safeguard-20b
Pricing unavailable
OpenAI
gpt-oss-safeguard-20b
Pricing unavailable

FAQ