openai/gpt-oss-20b · Hugging Face

Fields

id

sources/gpt-oss-20b-hf-card

URL

https://huggingface.co/openai/gpt-oss-20b

Kind

repo

Accessed

2026-10-19

notes

Read https://huggingface.co/openai/gpt-oss-20b for the benchmark figures on models/gpt-oss-20b. As recorded by the reader: HF model card has no score table; it links to the official model card paper (arXiv:2508.10925). Values are Table 3, gpt-oss-20b, reasoning level 'high', no-tools rows. The paper also publishes with-tools variants (AIME 2024 96.0, AIME 2025 98.7, GPQA Diamond 74.2) and low/medium levels, not reported here. No MMLU-Pro, HumanEval, IFEval or LiveCodeBench entries for the 20B model.

Notes

Read https://huggingface.co/openai/gpt-oss-20b for the benchmark figures on models/gpt-oss-20b. As recorded by the reader: HF model card has no score table; it links to the official model card paper (arXiv:2508.10925). Values are Table 3, gpt-oss-20b, reasoning level 'high', no-tools rows. The paper also publishes with-tools variants (AIME 2024 96.0, AIME 2025 98.7, GPQA Diamond 74.2) and low/medium levels, not reported here. No MMLU-Pro, HumanEval, IFEval or LiveCodeBench entries for the 20B model.

Cross-references

This record references no other records.