Open Systems, Blockchain & Business Daily

OSBBD

AI, money and the software behind them.

gpt-oss-20b

gpt-oss-20b is an open-weight language model from OpenAI, released on August 5, 2025. It is a mixture-of-experts model with 21B total parameters, of which 3.6B are active for each token. OpenAI lists a context window of 128K tokens. The model takes text input.

Developer
OpenAI
Released
August 5, 2025
Total parameters
21B
Active parameters per token
3.6B
Architecture
MoE
Context window
128K
Input
Text
License
Apache 2.0
Commercial use
Yes
Official weights, GB
13.8 GB
Hugging Face repository
huggingface.co
Checked on

The weights are released under Apache 2.0, which allows commercial use.

The official weights on Hugging Face take about 13.8 GB. Running the model needs at least that much memory across GPUs and system RAM, plus room for the context cache; quantized versions need less.

Total parameters, billions
21B
Active parameters, billions
3.6B
Context window, tokens
128K
License type
Apache 2.0
License conditions
Apache 2.0, with a short usage policy in the repository. No user or revenue limits.
Weights precision
MXFP4 + BF16
License text
huggingface.co
Sources
PageURL
Hugging Face model cardhttps://huggingface.co/openai/gpt-oss-20b/raw/main/README.md
Hugging Face weights indexhttps://huggingface.co/openai/gpt-oss-20b/raw/main/model.safetensors.index.json
openai.comhttps://openai.com/index/introducing-gpt-oss/
Hugging Face APIhttps://huggingface.co/api/models/openai/gpt-oss-20b?expand=safetensors
OSBBD

What are you looking for?

Search by headline, topic or keyword.