LLM
gpt-oss-20b
OpenAI's 21B open-weight MoE text model; base of gpt-oss-safeguard-20b and GPT-OSS-Swallow-20B.
Levelmodel
ClassLLM
Parameters21B
Context131,000
Released5 Aug 2025
Opennessopen weights permissive
Runs onconsumer GPU
Statusactive
Benchmark Results
maintainer-run
SEA-HELMSoutheast Asian and Indonesian
Checkpoint gpt-oss-20b as listed, quantization mxfp4; release view (data last updated 18 Sep 2026); score is the average of 2000 bootstrapsmaintainer-run
68.38 95% CI [-1.34, +1.35]
04080
Score in score. Higher is better.
SEA-HELM leaderboard, 58 models, Indonesian (ID)
Checkpoint gpt-oss-20b as listed, quantization mxfp4; release view (data last updated 18 Sep 2026); score is the average of 2000 bootstrapsmaintainer-run
60.64 95% CI [-0.64, +0.63]
04080
Score in score. Higher is better.
SEA-HELM leaderboard, 58 models, SEA average
Where this comes from
Every score names who ran it: vendor-reported, maintainer-run or independent. A rank is shown with the population it was ranked in (gap rule 3).
Lineage
continued pretrainfinetuneproduct of
GPT-OSS-Swallow-20B is a child (continued pretrain)medium confidencenotesgpt-oss-safeguard-20b is a child (finetune)high confidencemodel cardgpt-oss is the parent (product of)high confidencenotesWhere this comes from
Parent and child relations from model cards, Hugging Face base_model fields and the genealogy tables of the research notes. The tree shows two generations each way, 5 models. Confidence and evidence are the notes' own.
R&D Topics
Identity
Where this comes from
Developer and dates from the research notes. Release date is the vendor's public announcement; weights date is when open weights were published.
Architecture and Size
- Architecture
- MoE
- Total parameters
- 21B sources
- Active parameters
- 3.6B
- Checkpoint parameters
- –
- Context length
- 131,000 tokens
- Input
- text
- Output
- text
Where this comes from
Parameter counts as the developer states them. Where the checkpoint holds a different total, both are shown (register A16).
Openness and Access
- Openness
- open weights permissive
- Licence
- Apache License 2.0 sources
- Access routes
- open weights download
Where this comes from
The licence file governs over registry tags and aggregator labels.
Deployment
- Runs on
- consumer GPU
- Footprint
- 16 GB consumer GPU
- Runtimes
- not in the notes
Where this comes from
Deployment tiers and runtimes as named in the notes. Footprint is the notes' figure, not a measurement by AURION.
Indonesia
- Indonesian support
- not stated
- Regional languages
- not in the notes
- Jakarta hosting
- unknown
- Languages
- not in the notes
Where this comes from
Indonesia fields are carried on every model; unknown is an allowed value.
Industry Use Cases
No industry use case links this model.
Links
DeveloperOpenAIfrontier lab, USA
OpenAIfrontier lab, USAModelgpt-oss-20bOpenAI, LLM
gpt-oss-20bOpenAI, LLMR&D TopicsSmall Language Models (SLM)Efficiency Architectures (MoE, Sparse Attention, Long Context)Open-Weight Frontier Releases as Research SubstrateRegional or National Foundation Model Training (Continued Pretraining at 7B to 70B)
Use CasesIndonesian and Regional-Language LLMsPublic SectorIndustrial Copilots and Maintenance AssistantsIndustrialTelecom Domain LLMs and BenchmarksTelecomBanking Customer Service AssistantsFinance+4 moreUse cases of models in the same R&D topic; none is recorded for this model itself.
Who Runs It
PT Pertamina Geothermal EnergyIn production, 14 Jul 2026
PT Amman Mineral Nusa Tenggara (AMMAN)In production, 8 Jul 2026
PT PegadaianIn production, 5 Jun 2026
Bank Central Asia (BCA)In production, 16 Mar 2026+28 moreDeployments of these use cases; none names this model.
PT Pertamina Geothermal EnergyIn production, 14 Jul 2026PT Amman Mineral Nusa Tenggara (AMMAN)In production, 8 Jul 2026PT PegadaianIn production, 5 Jun 2026Bank Central Asia (BCA)In production, 16 Mar 2026+28 moreDeployments of these use cases; none names this model.Models 3
Organisations 1
R&D topics 4
Open-Weight Frontier Releases as Research SubstrateUsing released weights as the base for downstream research.used bySmall Language Models (SLM)Models under about 10B parameters that fit consumer devices.exemplarEfficiency Architectures (MoE, Sparse Attention, Long Context)Reducing training and inference cost per capability.produced byRegional or National Foundation Model Training (Continued Pretraining at 7B to 70B)Pretraining or continued pretraining a mid-size model to serve a national or regional lan…used by
Sources
- openai/gpt-oss-120b , Hugging Face huggingface.co https://huggingface.co/openai/gpt-oss-120b
- huggingface.co/openai https://huggingface.co/openai
- openai/gpt-oss-safeguard-20b , Hugging Face huggingface.co https://huggingface.co/openai/gpt-oss-safeguard-20b
- MLCommons MLPerf Training Benchmark mlcommons.org https://mlcommons.org/benchmarks/training/
- meta-llama/Llama-3.1-8B-Instruct , Hugging Face huggingface.co https://huggingface.co/meta-llama/Llama-3.1-8B-Instruct
- openai.com/index/introducing-gpt-oss https://openai.com/index/introducing-gpt-oss/
- gigazine.net/gsc_news/en/20250806-openai-gpt-oss https://gigazine.net/gsc_news/en/20250806-openai-gpt-oss
- huggingface.co/api/models/openai/gpt-oss-20b https://huggingface.co/api/models/openai/gpt-oss-20b
- github.com/openai/gpt-oss https://github.com/openai/gpt-oss
To cite this page: Project AURION, AI Use-Cases Research & Innovation, "Gpt-Oss-20b", https://aurion.semai.id/models/openai-gpt-oss-20b/, records of October 2026.