skip to content
gigarouter gigarouter
models / embeddings · coming soon

DenseOn Unsupervised

lightonai/DenseOn-unsupervised

published Mar 2026 · updated Apr 2026

DenseOn Unsupervised is an embed model that provides a strong intermediate dense checkpoint from the first stage of the DenseOn pipeline, using unsupervised contrastive pre-training on filtered query-document pairs.

est. price
~$0.008
/ 1M tokens · estimated, set at launch
API providers
0
downloads / mo
140
license
apache-2.0

specs

TaskEmbeddings / Dense Retrieval
ArchitectureModernBERT-base with CLS token pooling
Parameters149M
LicenseApache 2.0

best for

FAQ

What is DenseOn Unsupervised best used for?

It is intended as a starting point for supervised fine-tuning, knowledge distillation, or research ablations, not for direct production retrieval.

How does DenseOn Unsupervised compare to the full DenseOn model?

DenseOn Unsupervised (49.05 BEIR NDCG@10) is the pre-training-only checkpoint; the full DenseOn (56.20) adds supervised fine-tuning with mined hard negatives and is recommended for production.

What is the model architecture and size?

It uses a ModernBERT-base backbone with 149M parameters, 768-dimensional embeddings, and CLS token pooling with a 512 token limit.

What is the license for DenseOn Unsupervised?

Apache 2.0.

How do I call this model via the gigarouter API?

Use the gigarouter OpenAI-compatible endpoint with your API key, specifying the model ID lightonai/DenseOn-unsupervised.

not yet live

We're benchmarking and onboarding DenseOn Unsupervised as a hosted, OpenAI-compatible API. Sign in for free credit and be ready when it lands, or tell us you want it and we'll prioritize it.

related embeddings models

compare all →