gpt-oss-safeguard-20b
A
AI Socratic
OpenAI released gpt-oss-safeguard-20b on September 18, 2025. The registry currently records it as open weights.
Model publication
OpenAI
gpt-oss-safeguard-20bOpen the source for gpt-oss-safeguard-20b
Open weights
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 & safety labeling. Learn more about this model in OpenAI's gpt-oss-safeguard [user guide](https://cookbook.openai.com/articles/gpt-oss-safeguard-guide).
- Licence
- apache-2.0
- Architecture
- GptOssForCausalLM
- Parameters
- 21.5B
- Context
- 131.1K tokens
- Artifact format
- safetensors
- Quantization
- mxfp4 4-bit
- Download size
- 13 GB
Hardware to run it (inference, estimated)
13 GB VRAM minimum · 16 GB recommended · 16 GB system RAM · 13 GB storage
Estimated from parameter count and stored precision; verify against the selected runtime and context length.
Benchmarks
No benchmark observation recorded for this model.
About the Authors
A