Skip to content

OpenAI · Chat

GPT OSS Safeguard 20B

OpenAI 🇺🇸ChatGPT familyOpen weights. One of its sources says the weights are publicly available.
From NanoGPT's listing:
gpt-oss-safeguard is a first open weight reasoning model specifically trained for safety classification tasks to help classify text content based on customizable policies. As a fine-tuned version of gpt-oss, gpt-oss-safeguard is designed to follow explicit written policies that you provide. This enables bring-your-own-policy Trust & Safety AI, where your own taxonomy, definitions, and thresholds guide classification decisions. Well crafted policies unlock gpt-oss-safeguard's reasoning capabilities, enabling it to handle nuanced content, explain borderline decisions, and adapt to contextual factors.
Max context
131K
Listings
23
Providers
17

Evidence and sources

🤗 openai/gpt-oss-safeguard-20bGood value (61.7). Cheaper per task than its benchmark results predict. From 1 benchmark with measured task costs. Little data behind it, so treat it as a rough estimate.How Value for money worksReleased 18 Sep 2025Maker HQ: US (according to OpenAI's own website)
All rankings: #170 on NanoGPT recently #210 on OpenRouter today #235 on OpenRouter this week #273 on OpenRouter this monthSources: AWS Bedrock Runtime data, Models.dev data, AI/ML API data, NanoGPT, Hugging FacePrimary data from AWS Bedrock Runtime from Models.dev: weights availability from AI/ML API: primary mode from NanoGPT: description from Hugging Face: release date

Advantages and trade-offs

Compared with 10 representative chat models

Advantages

Open weights available

The weights are public. 4 of the 10 peers that say either way have open weights.

We only make a claim when there's enough data on models of the same type. Updated 22 Nov, 06:07 UTC.Prices come from the model's maker where they sell it, otherwise their official partner, otherwise the market median. This comparison uses 7 maker's prices.

Input & output

as stated by the model's providers
Text
Text

Capabilities

Supported operations and provider-stated features

ReasoningFunction callingStructured outputPrompt cachingWeb searchBatch API

Listings

 OpenAIgpt-oss-safeguard-20bSource: GenAI Prices dataNot published by the provider$0.075cache read $0.037$0.30——
 Amazon Web Servicesopenai.gpt-oss-safeguard-20bSources: AWS Bedrock Runtime data, LiteLLM data, Models.dev data131K$0.07batch $0.03$0.20batch $0.10——
 Eden AIInference by Amazon Web Servicesamazon/openai.gpt-oss-safeguard-20bSource: Models.dev data128K$0.07$0.20——
 Hugging FaceInference by Amazon Web Servicesopenai/gpt-oss-safeguard-20bSource: Hugging Face Inference Providers data131K$0.07$0.20——
 Kiloopenai/gpt-oss-safeguard-20bSources: Kilo AI Gateway data, Models.dev data131K$0.07$0.20——
 OpenRouterInference by Amazon Web Servicesopenai/gpt-oss-safeguard-20b · us-east-2Source: OpenRouter data docs131K$0.07$0.20233.5362.5ms
 Vercelopenai/gpt-oss-safeguard-20bSources: Models.dev data, Vercel AI Gateway data128K$0.07$0.20——
 Merge Gatewayopenai/gpt-oss-safeguard-20bSource: Models.dev data4K$0.07$0.20——
 NanoGPTInference by KitaniRegions: GLBopenai/gpt-oss-safeguard-20b · mxfp4Source: NanoGPTNot published by the provider$0.0709cache read $0.0354$0.294—32,926ms
 Nous Portalopenai/gpt-oss-safeguard-20bSource: Nous Portal Inference Gateway data131K$0.075cache read $0.0375$0.30——
Show all 19 listings

Showing 10 of 19: the maker's own listing first, then the cheapest per 1M input tokens from each provider.

Also sold by Aimlapi, Fireworks AI, Groq, Lowrouter.

4 listings without a price are hidden.

These providers advertise this model. Researchers have found some resellers serving a different, cheaper model, so see how to check whether a provider meets your needs.

Benchmarks

Named effort levels run from lowest to highest. Select a result to see its source details.

Reasoning

Data credits: SnakeBench (Greg Kamradt / ARC Prize Foundation), MIT (licence)

Also known as: amazon/openai.gpt-oss-safeguard-20b, amazon/openai.gpt-oss-safeguard-20b@us, bedrock/ap-northeast-1/openai.gpt-oss-safeguard-20b, bedrock/ap-south-1/openai.gpt-oss-safeguard-20b, bedrock/eu-south-1/openai.gpt-oss-safeguard-20b, bedrock/eu-west-1/openai.gpt-oss-safeguard-20b, bedrock/eu-west-2/openai.gpt-oss-safeguard-20b, bedrock/sa-east-1/openai.gpt-oss-safeguard-20b, bedrock_mantle/ap-northeast-1/openai.gpt-oss-safeguard-20b, bedrock_mantle/ap-south-1/openai.gpt-oss-safeguard-20b, bedrock_mantle/ap-southeast-2/openai.gpt-oss-safeguard-20b, bedrock_mantle/ap-southeast-3/openai.gpt-oss-safeguard-20b, bedrock_mantle/eu-central-1/openai.gpt-oss-safeguard-20b, bedrock_mantle/eu-north-1/openai.gpt-oss-safeguard-20b, bedrock_mantle/eu-south-1/openai.gpt-oss-safeguard-20b, bedrock_mantle/eu-west-1/openai.gpt-oss-safeguard-20b, bedrock_mantle/eu-west-2/openai.gpt-oss-safeguard-20b, bedrock_mantle/openai.gpt-oss-safeguard-20b, bedrock_mantle/sa-east-1/openai.gpt-oss-safeguard-20b, fireworks_ai/accounts/fireworks/models/gpt-oss-safeguard-20b, groq/openai/gpt-oss-safeguard-20b, openai.gpt-oss-safeguard-20b, openai/gpt-oss-safeguard-20b

Like this one. Tap or hover any dotted value, label or icon to see where it comes from and how it is measured. open details when tapped or hovered. How this data works