gpt-oss-120b & gpt-oss-20b Model Card

Fields

id

sources/model-fr-openai-gpt-oss-card

URL

https://huggingface.co/openai/gpt-oss-120b

Publisher

OpenAI

Kind

spec-sheet

Published

2025-08-05

Accessed

2026-10-03

notes

States 117B total / 5.1B active for the 120b and 21B total / 3.6B active for the 20b. MXFP4 quantization of the MoE weights is the post-training format, which is what makes the 120b fit on one 80GB device.

Notes

States 117B total / 5.1B active for the 120b and 21B total / 3.6B active for the 20b. MXFP4 quantization of the MoE weights is the post-training format, which is what makes the 120b fit on one 80GB device.

Cross-references

This record references no other records.