GPT OSS Safeguard 20B: API pricing, providers and specs
GPT OSS Safeguard 20B is a moderation model by OpenAI. 5 providers serve it, from $0.20 to $0.317 /1M; the lowest price is at Vercel AI Gateway.
A newer version is out: GPT OSS Safeguard 120B · released Sep 18, 2025
How GPT OSS Safeguard 20B compares
Against the 10 moderation models on the site priced the same way.
- Mid-priced: its lowest price is $0.20 /1M against a median of $0.18; 55% of comparable models cost less.
- 5 providers serve it; 82% of comparable models have fewer.
- Its weights are open under the Apache 2.0 license: commercial use allowed.
GPT OSS Safeguard 20B prices by provider
Every provider's list price, read from its own price page or API. A price marked “from” changes with context length, resolution or mode.
| Provider | Input | Output | Prompt caching | Context |
|---|---|---|---|---|
| Vercel AI Gateway | $0.07/1M | $0.20/1M | — | 128K |
| $0.07/1M | From$0.20/1M | — | — | |
| Groq | $0.075/1M | $0.30/1M | — | — |
| OpenRouter | $0.075/1M | From$0.30/1M | $0.037/1M | 131K |
| AI/ML API | $0.079/1M | $0.317/1M | $0.04/1M | — |
- Vercel AI Gateway$0.20
- Amazon BedrockFrom$0.20
- Groq$0.30
- OpenRouterFrom$0.30
- AI/ML API$0.317
Specs
- Type
- Moderation
- Maker
- OpenAI
- Inputs
- Text
- Outputs
- Text
- Context
- 131K≈ 200 pages of text
- Max output
- 66K
- Capabilities
- Reasoning
- Released
- Sep 18, 2025
- License
- Apache 2.0commercial use allowed
- Open weights
- Yes, downloadable
- Parameters
- 22B
Alternatives to GPT OSS Safeguard 20B
Same kind of model from other makers, at a similar price.
| Model | Lowest price | Context | Arena | Providers |
|---|---|---|---|---|
| GPT OSS Safeguard 20Bthis model | $0.20/1M | 131K | — | 5 |
| $0.20/1M | 131K | — | 3 | |
| $0.26/1M | — | — | 2 | |
| $0.18/1M | 164K | — | 6 | |
| $0.20/1M | — | — | 1 |
More from OpenAI
GPT OSS Safeguard 20B FAQ
The lowest price for GPT OSS Safeguard 20B is $0.20 /1M, at Vercel AI Gateway.
Vercel AI Gateway, at $0.20 /1M. 5 providers serve GPT OSS Safeguard 20B; the table above lists every one with its price.
GPT OSS Safeguard 20B reads up to 131K tokens per request and writes up to 66K.
Yes. GPT OSS Safeguard 20B's weights can be downloaded from its maker's Hugging Face page, under the Apache 2.0 license.
About 22B, as the safetensors metadata of its official Hugging Face repo states.
5 API providers serve GPT OSS Safeguard 20B. Compare their prices above, or open a provider to see everything it offers.
Can't find what you're looking for? Contact us