gpt-oss-20b is an open-weight 21B parameter model released by OpenAI under the Apache 2.0 license. It uses a Mixture-of-Experts (MoE) architecture with 3.6B active parameters per forward pass, optimized for lower-latency inference. The model is trained in OpenAI’s Harmony response format and supports reasoning level configuration, fine-tuning, and agentic capabilities including function calling, tool use, and structured outputs.
Modalities
In / out price
$0.035 / $0.161per 1M
Context
131K
Max output
131K
Updated
Nov 5, 2026
Providers
4 routesCompare providers by price, privacy, and speed.
| Private | $0.035 | $0.161 | — | 131K | 131K | — | — | |
Amazon Bedrock | Private | $0.081 | $0.35 | — | 128K | 16K | — | — |
Cheaper Inference | Anonymous | $0.065 | $0.25 | $0.065 | 128K | 128K | — | — |
Surplus Intelligence | Anonymous | $0.042 | $0.18 | $0.042 | 128K | 16K | — | — |
Privacy
2 / 4Private routing
Private tier
- Anonymous
- Private
- TEE
- E2EE
Prompt and response content is not retained after the request. Request content is visible to the inference runtime while it is being processed. Learn more
Features
3Uptime
FAQ
Common questions about OpenAI GPT OSS 20B.