Why did we open-source our inference engine? Read the post

lightonai/Reason-ModernColBERT (Score)

Reason-ModernColBERT is a late interaction model trained on the reasonir-hq dataset. It achieves extremely competitive performance on the BRIGHT benchmark aimed at evaluating reasoning-intensive retrieval performance, outperforming all existing models up to 7B (more than 45 times its size) and ev...

Overview

Architecture
ModernBERT
Parameters
149M
Tasks
Encode, Score
Outputs
Multi-Vec
Dimensions
Multi-Vec: 128
Max Sequence Length
8,192 tokens
License
cc-by-nc-4.0
Languages
en

Benchmarks

AskUbuntuDupQuestions

technology reranking en

Duplicate question detection from AskUbuntu

Corpus: 6,743 Queries: 360
Quality
ndcg at 10 0.6520
map at 10 0.4972
mrr at 10 0.7335
Reference →

CMedQAv1-reranking

general reranking en

Quality
ndcg at 10 0.4245
map at 10 0.3554
mrr at 10 0.4450

CMedQAv2-reranking

general reranking en

Quality
ndcg at 10 0.4482
map at 10 0.3825
mrr at 10 0.4718

MMarcoReranking

general reranking zh

Multilingual MARCO passage reranking (Chinese)

Quality
ndcg at 10 0.1861
map at 10 0.1491
mrr at 10 0.1551
Reference →

Self-hosted inference for search & document processing

Cut API costs by 50x, boost quality with 85+ SOTA models, and keep your data in your own cloud.

Github 1.9K

Contact us

Tell us about your use case and we'll get back to you shortly.