OpenAI

GPT OSS Safeguard 20B

Generally availableOpen weights128K contextReasoning

Summary

GPT OSS Safeguard 20B is an open-weight language model from OpenAI, released on 29 Oct 2025. It accepts text and generates text, with a 128K-token context window and up to 16K output tokens per response. As of 18 Nov 2026, the lowest listed API price is $0.07 per million input tokens and $0.20 per million output tokens (Amazon Bedrock), across 4 providers we track.

Safety model for policy screening, moderation, and risk-aware routing workflows

Last verified 18 Nov 2026. Every figure on this page links to its source.

Specifications

DeveloperOpenAI
Release date29 Oct 2025 [models.dev]
StatusGenerally available [models.dev]
WeightsOpen weights [Hugging Face]
Licenseapache-2.0 [Hugging Face]
Parameters21.5B [Hugging Face]
Context window128K (128,000 tokens) [models.dev]
Max output16K tokens [models.dev]
InputText [models.dev]
OutputText [models.dev]
Reasoning modeYes [models.dev]
Tool callingYes [models.dev]
Structured outputYes [models.dev]
Hugging Faceopenai/gpt-oss-safeguard-20b [Hugging Face]

GPT OSS Safeguard 20B API pricing by provider

USD per million tokens. Sorted by input price.

ProviderInputOutputCached inputContextSource
Amazon Bedrock
openai.gpt-oss-safeguard-20b
$0.07$0.20–128Kmodels.dev, 18 Nov 2026
Vercel AI Gateway
openai/gpt-oss-safeguard-20b
$0.07$0.20–128Kmodels.dev, 18 Nov 2026
Groq
openai/gpt-oss-safeguard-20b
$0.07$0.30$0.04131Kmodels.dev, 18 Nov 2026
OpenRouter
openai/gpt-oss-safeguard-20b
$0.07$0.30$0.04131KOpenRouter, 18 Nov 2026

Running GPT OSS Safeguard 20B locally

Estimated memory, from the parameter count and the architecture in config.json on Hugging Face. Includes about 0.8 GB of runtime overhead.

QuantizationWeightsTotal, 8K contextTotal, 32K context
Q4_K_M (4-bit)13 GB14 GB15 GB
Q8_0 (8-bit)23 GB24 GB24 GB
FP16 / BF16 (16-bit)43 GB44 GB45 GB

At 4-bit with an 8K conversation, GPT OSS Safeguard 20B needs about 14 GB, so it fits on a graphics card with 16 GB or more, such as the NVIDIA GeForce RTX 4060 Ti 16 GB. Check your own hardware, or see Best GPU for local LLMs on our sister site AIPCs.

Frequently asked questions

What is the context window of GPT OSS Safeguard 20B?

GPT OSS Safeguard 20B has a context window of 128,000 tokens (128K) and can generate up to 16,384 tokens in a single response, according to models.dev.

How much does the GPT OSS Safeguard 20B API cost?

GPT OSS Safeguard 20B is listed from $0.07 per million input tokens and $0.20 per million output tokens (Amazon Bedrock). Prices last verified 18 Nov 2026.

Is GPT OSS Safeguard 20B open source?

GPT OSS Safeguard 20B is an open-weight model: its weights are published on Hugging Face as openai/gpt-oss-safeguard-20b under the apache-2.0 license. Check the license terms before commercial use.

When was GPT OSS Safeguard 20B released?

OpenAI released GPT OSS Safeguard 20B on 29 Oct 2025, according to models.dev.

Which providers offer GPT OSS Safeguard 20B?

We track GPT OSS Safeguard 20B on 4 providers: Amazon Bedrock, Vercel AI Gateway, Groq, OpenRouter.

How much VRAM does GPT OSS Safeguard 20B need?

About 14 GB at 4-bit (Q4_K_M) with an 8K-token conversation, and about 24 GB at 8-bit. The 4-bit model fits on a graphics card with 16 GB or more, such as the NVIDIA GeForce RTX 4060 Ti 16 GB. These are estimates from the model's published parameter count and architecture.

Similar models from other labs

Comparable on output type, weights, release date, price and context window. Prices are the lowest listed, USD per million tokens.

ModelLabReleasedContextInput $/1MOutput $/1MWeights
Ministral 3 3BMistral AI2 Dec 2025131K$0.10$0.10Open
Nvidia Nemotron Nano 9B V2NVIDIA18 Aug 2025131K$0.06$0.23Open
UI-TARS 7BBytedance22 Jul 2025128K$0.10$0.20Open
Granite 4.0 H MicroIBM2 Oct 2025131K$0.02$0.11Open
ERNIE 4.5 21B A3BBaidu30 Jun 2025120K$0.07$0.28Open
Ministral 3 8BMistral AI2 Dec 2025262K$0.15$0.15Open

Change history

  • Name changed from gpt-oss-safeguard-20b to GPT OSS Safeguard 20B
  • Max output changed from 16K to 16K
  • Context window changed from 131K to 128K
  • Name changed from GPT OSS Safeguard 20B to gpt-oss-safeguard-20b
  • Max output changed from 66K to 16K
  • Context window changed from 128K to 131K
  • Max output changed from 16K to 66K
  • Max output changed from 66K to 16K
  • Context window changed from 131K to 128K
  • Max output changed from 16K to 66K
  • Context window changed from 128K to 131K
  • Name changed from gpt-oss-safeguard-20b to GPT OSS Safeguard 20B
  • Max output changed from 16K to 16K
  • Max output changed from 66K to 16K
  • Name changed from GPT OSS Safeguard 20B to gpt-oss-safeguard-20b
  • Context window changed from 131K to 128K
  • Name changed from gpt-oss-safeguard-20b to GPT OSS Safeguard 20B
  • Context window changed from 128K to 131K
  • Max output changed from 16K to 66K
  • Status changed from preview to active
  • Context window changed from 131K to 128K
  • Max output changed from 66K to 16K
  • Name changed from Safety GPT OSS 20B to gpt-oss-safeguard-20b
  • Name changed from GPT OSS Safeguard 20B to Safety GPT OSS 20B
  • Status changed from active to preview

Sources