language encoderfamily
DeBERTa
Encoder with disentangled attention over content and position; version 3 uses ELECTRA-style pre-training and is a common base for classification and NLI models.
Levelfamily
Classlanguage encoder
Parametersnot in the notes
Contextnot in the notes
Released5 Jun 2020
Opennessopen weights permissive
Runs onnot in the notes
Statusactive
Lineage
DeBERTa-v3 base belongs to this familylanguage encoder18 Nov 2021DeBERTa-v3 large belongs to this familylanguage encoder18 Nov 2021Where this comes from
Parent and child relations from model cards, Hugging Face base_model fields and the genealogy tables of the research notes. Confidence and evidence are the notes' own.
Identity
- Developer
- Microsoft AI
- Family
- not in the notes
- Released
- 5 Jun 2020 sourcesannouncement page
- Weights published
- –
- Record checked
- 12 Oct 2026
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
- dense transformer
- Total parameters
- – sources
- Active parameters
- –
- Checkpoint parameters
- –
- Note
- v1, v2 and v3 lines; the multilingual mDeBERTa-v3 base is a separate AURION record (microsoft-mdeberta-v3-base).
- Context length
- not in the notes
- 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
- MIT License sources
- Access routes
- open weights download
Where this comes from
The licence file governs over registry tags and aggregator labels.
Deployment
- Runs on
- not in the notes
- Footprint
- not in the notes
- 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
- Model card languages: en
Where this comes from
Indonesia fields are carried on every model; unknown is an allowed value.
Benchmark Results
The notes record no benchmark result for this model.
Industry Use Cases
No industry use case links this model.
Links
DeveloperMicrosoft AIbig tech, USA
Microsoft AIbig tech, USAModelDeBERTaMicrosoft AI, language encoder
DeBERTaMicrosoft AI, language encoderR&D TopicsPretrained Bidirectional Text Encoders
Use CasesIndonesian NLU (Classification, NER)Public SectorEmbeddings and Retrieval for SEA LanguagesPublic SectorUse cases of models in the same R&D topic; none is recorded for this model itself.
Who Runs Itnone recordedDeployments of these use cases; none names this model.
Models 2
Organisations 1
Sources
- arxiv.org/abs/2006.03654 https://arxiv.org/abs/2006.03654
- github.com/microsoft/DeBERTa https://github.com/microsoft/DeBERTa
- kaggle.com/models/keras/deberta_v3 https://www.kaggle.com/models/keras/deberta_v3
- paperswithcode.com/method/deberta https://paperswithcode.com/method/deberta
- huggingface.co/microsoft/deberta-v3-base https://huggingface.co/microsoft/deberta-v3-base
- huggingface.co/microsoft/deberta-v3-large https://huggingface.co/microsoft/deberta-v3-large
To cite this page: Project AURION, AI Use-Cases Research & Innovation, "DeBERTa", https://aurion.semai.id/models/microsoft-deberta/, records of October 2026.