Skip to content

EthenEthenEthen

LoginPlatformTry Ethen

Open Source Model Profile · nreimers

MiniLM-L6-H384-uncased

MiniLM-L6-H384-uncased is a BERT feature-extraction model from nreimers. Its model card describes it as a 6-layer version of microsoft/MiniLM-L12-H384-uncased.

Publisher
nreimers
Task
feature-extraction
Model type
bert
License
mit
Library
transformers
Publication status
Accepted · not indexed

Model overview

MiniLM-L6-H384-uncased is published by nreimers as a feature-extraction model. The captured configuration identifies BertModel with model type bert, and card data records MIT licensing. According to the model card, it is a 6-layer version of microsoft/MiniLM-L12-H384-uncased made by keeping only every second layer.

Recorded capabilities

Documented 6-layer derivation

According to the model card, this is a 6-layer version of microsoft/MiniLM-L12-H384-uncased made by keeping only every second layer.

BERT feature-extraction record

The captured configuration identifies BertModel with model type bert for the feature-extraction pipeline.

MIT license

Card data records MIT licensing, and hub tags include a license:mit entry.

Use cases in the source record

  • Feature-extraction and sentence-embedding workflows using the captured BERT configuration with Transformers support.
  • Compact-encoder experiments where the publisher describes a 6-layer MiniLM alternative to the 12-layer source model.

Limitations and unknowns

  • No parameter count was extracted from this record.
  • No evaluation results, training hyperparameters, or context-window value were extracted.
  • Provider state is historical snapshot data and should be refreshed before being presented as current.
  • Derivation detail comes from a brief publisher model card and was not independently verified by Ethen.

Source and provenance

Source: nreimers/MiniLM-L6-H384-uncased

Captured: Unknown. Processed: 2026-09-07T19:34:53.416293+00:00.

MiniLM: 6 Layer Version This is a 6 layer version of microsoft/MiniLM-L12-H384-uncased by keeping only every second layer.

F001F002F003F004F005F006F007F008F009F010