← Catalog
lightonai/GTE-ModernColBERT-v1 Add to compare Open comparison →
Primitive: /encode · Encode ·
ModernBERT
This is a multi-vector (ColBERT-style late interaction) embedding model trained on the ms-marco-en-bge-gemma dataset. It maps sentences & paragraphs to sequences of 128-dimensional dense vectors and can be used for semantic textual similarity using the MaxSim operator.
Long context Multi-vector
View on Hugging Face → Fine-tuned from Alibaba-NLP/gte-modernbert-base
Overview
Hardware:
L4 RTX-PRO-6000 — drives latency, throughput & cost
Size 149M params Tasks /encode · /score License apache-2.0 Latency 99 ms Throughput 28.0K tok/s Cost $0.0079 /1M tok
Cost is approximate — computed from list GPU prices; your actual price depends on the provider you deploy SIE with.
Embedding Output types Multi-Vec Dimensions multivector: 128 Max sequence length 8,192 Inputs text
Benchmarks Duplicate question retrieval from StackExchange Physics
Corpus: 38,314 Queries: 1,039
Quality
map at 10 0.3410
mrr at 10 0.3905
ndcg at 10 0.3887
Performance L4 b1 c16
Corpus 21.7K tok/s
Corpus p50 88.1ms
Query 2.5K tok/s
Query p50 68.2ms
Reference →
Code search with natural language queries
Corpus: 6,267 Queries: 500
Quality
ndcg at 10 0.3301
map at 10 0.2512
mrr at 10 0.2343
Performance L4 b1 c16
Corpus 7.4K tok/s
Corpus p50 84.4ms
Query 462 tok/s
Query p50 76.3ms
Reference →
Financial opinion mining and question answering
Corpus: 57,599 Queries: 648
Quality
map at 10 0.3126
mrr at 10 0.4641
ndcg at 10 0.3833
Performance L4 b1 c16
Corpus 18.9K tok/s
Corpus p50 106.6ms
Query 2.4K tok/s
Query p50 71.9ms
Reference →
Question answering on consumer contracts
Corpus: 153 Queries: 396
default
Quality
ndcg at 10 0.7773
map at 10 0.7300
mrr at 10 0.7321
Performance L4 b1 c16
Corpus 42.9K tok/s
Corpus p50 192.4ms
Query 3.6K tok/s
Query p50 70.2ms
long_context
Quality
ndcg at 10 0.8188
map at 10 0.7710
mrr at 10 0.7718
Reference →
Biomedical literature search from NutritionFacts.org
Corpus: 3,593 Queries: 323
Quality
map at 10 0.1392
mrr at 10 0.5808
ndcg at 10 0.3618
Performance L4 b1 c16
Corpus 35.9K tok/s
Corpus p50 101.3ms
Query 1.7K tok/s
Query p50 45.7ms
Reference →
Smaller subset of the FiQA financial QA dataset
default
Quality
ndcg at 10 0.6032
map at 10 0.5264
mrr at 10 0.6721
Reference →
Citation prediction, document classification, and recommendation for scientific papers
Corpus: 25,656 Queries: 1,000
Quality
map at 10 0.0934
mrr at 10 0.2875
ndcg at 10 0.1608
Performance L4 b1 c16
Corpus 30.1K tok/s
Corpus p50 96.3ms
Query 2.1K tok/s
Query p50 68.6ms
Reference →
Scientific claim verification using research literature
Corpus: 5,183 Queries: 300
Quality
ndcg at 10 0.7600
map at 10 0.7179
mrr at 10 0.7302
Performance L4 b1 c16
Corpus 51.0K tok/s
Corpus p50 82.9ms
Query 3.8K tok/s
Query p50 72.0ms
Reference →
Programming question answering from Stack Overflow
Corpus: 19,931 Queries: 1,994
Quality
ndcg at 10 0.6568
map at 10 0.6232
mrr at 10 0.6232
Performance L4 b1 c16
Corpus 26.0K tok/s
Corpus p50 127.7ms
Query 52.9K tok/s
Query p50 91.7ms
Reference →
