GPT-OSS-Safeguard
language modelOpenAI
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...
Key facts
| Made by | OpenAI |
|---|---|
| Type | language model |
| Released | October 29, 2025 |
| Reads at once | 131,072 tokens |
| Price (API, per million tokens) | $0.08 in · $0.3 out |
| Input → output | text → text |
| Sizes | 20B |
Questions about GPT-OSS-Safeguard
Who makes GPT-OSS-Safeguard?
GPT-OSS-Safeguard is made by OpenAI.
When was GPT-OSS-Safeguard released?
We first saw GPT-OSS-Safeguard listed on October 29, 2025.
How much does GPT-OSS-Safeguard cost?
Through OpenRouter it costs $0.08 per million input tokens and $0.3 per million output tokens. Prices change, so check the provider before you build on it.
How much text can GPT-OSS-Safeguard read at once?
Up to 131,072 tokens, roughly 98.3K words.
More from OpenAI
Whisper-SmallOpenAI · voice and audio model
Sep 26, 2022released
Whisper-Large-V3OpenAI · voice and audio model
Nov 7, 2023released
whisperOpenAI · voice and audio model
Sep 26, 2022released
Whisper-TinyOpenAI · voice and audio model
Sep 26, 2022released
Openai-Agents-Pythonopenai · developer tool
Mar 11, 2025released
Whisper-MediumOpenAI · voice and audio model
Sep 26, 2022released
Other language models
gpt2Openai Community · language model
Mar 2, 2022released
Qwen3.8Alibaba (Qwen) · language model
Aug 5, 2026released
Qwen3 VLAlibaba (Qwen) · language model
Sep 23, 2025released
Qwen2.5 VLAlibaba (Qwen) · language model
Jan 26, 2025released
optMeta · language model
May 11, 2022released
Qwen3.6Alibaba (Qwen) · language model
Apr 15, 2026released
Ollamaollama · language model
Jun 26, 2023released
pythiaEleutherAI · language model
Feb 8, 2023released
Modelcast collects this page automatically from public sources (OpenRouter). Facts are copied from those sources, not written by the maker. Not affiliated with OpenAI.