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

← Catalog

cross-encoder/ms-marco-MiniLM-L-6-v2

Open comparison →

Primitive: /score · Score · BERT

This model was trained on the MS Marco Passage Ranking task.

Overview

Hardware: — drives latency, throughput & cost

Size23M params
Tasks /score
Licenseapache-2.0
Languagesen
Latency46 ms
Throughput51.1K tok/s
Cost$0.0043 /1M tok

Cost is approximate — computed from list GPU prices; your actual price depends on the provider you deploy SIE with.

Scoring

Inputstext
Max sequence length512

Benchmarks

AskUbuntuDupQuestions

technology reranking en

Duplicate question detection from AskUbuntu

Corpus: 6,743 Queries: 360
Quality
ndcg at 10 0.6027
map at 10 0.4439
mrr at 10 0.6776
Performance L4 b1 c16
Query 948 tok/s
Query p50 362.8ms
Reference →

CMedQAv1Reranking

medical reranking zh

Chinese medical question answering reranking (v1)

Corpus: 100,000 Queries: 2,000
Quality
map at 10 0.0835
mrr at 10 0.1371
Reference →

CMedQAv2Reranking

medical reranking zh

Chinese medical question answering reranking (v2)

Corpus: 108,000 Queries: 4,000
Quality
map at 10 0.0926
mrr at 10 0.1425
Reference →

CQADupstackPhysicsRetrieval

scientific retrieval en

Duplicate question retrieval from StackExchange Physics

Corpus: 38,314 Queries: 1,039
Quality
map at 10 0.3669
mrr at 10 0.4301
ndcg at 10 0.4251
Reference →

CQADupstackPhysicsRetrieval (candidates: gte-multilingual-base, k=50)

scientific retrieval en

Duplicate question retrieval from StackExchange Physics

Performance L4 b1 c16
Query 44.3K tok/s
Query p50 44.6ms
Reference →

CosQA

technology retrieval en

Code search with natural language queries

Corpus: 6,267 Queries: 500
Quality
map at 10 0.1937
mrr at 10 0.2108
ndcg at 10 0.2618
Reference →

CosQA (candidates: gte-multilingual-base, k=50)

technology retrieval en

Code search with natural language queries

Performance L4 b1 c16
Query 20.5K tok/s
Query p50 43.6ms
Reference →

FiQA2018

finance retrieval en

Financial opinion mining and question answering

Corpus: 57,599 Queries: 648
Quality
map at 10 0.3160
mrr at 10 0.4782
ndcg at 10 0.3977
Reference →

FiQA2018 (candidates: gte-multilingual-base, k=50)

finance retrieval en

Financial opinion mining and question answering

Performance L4 b1 c16
Query 51.1K tok/s
Query p50 43.4ms
Reference →

LegalBenchConsumerContractsQA

legal retrieval en

Question answering on consumer contracts

Corpus: 153 Queries: 396
Quality
map at 10 0.6822
mrr at 10 0.6859
ndcg at 10 0.7409
Reference →

LegalBenchConsumerContractsQA (candidates: gte-multilingual-base, k=50)

legal retrieval en

Question answering on consumer contracts

Performance L4 b1 c16
Query 91.7K tok/s
Query p50 45.6ms
Reference →

MMarcoReranking

general reranking zh

Multilingual MARCO passage reranking (Chinese)

Quality
map at 10 0.0543
mrr at 10 0.0544
Performance L4 b1 c16
Reference →

NFCorpus

medical retrieval en

Biomedical literature search from NutritionFacts.org

Corpus: 3,593 Queries: 323
Quality
map at 10 0.2606
mrr at 10 0.5808
ndcg at 10 0.3591
Reference →

NFCorpus (candidates: gte-multilingual-base, k=50)

medical retrieval en

Biomedical literature search from NutritionFacts.org

Performance L4 b1 c16
Query 70.8K tok/s
Query p50 45.9ms
Reference →

NanoFiQA2018Retrieval

finance retrieval en

Smaller subset of the FiQA financial QA dataset

Performance L4 b1 c16
Query 7.5K tok/s
Query p50 388.1ms
Reference →

SCIDOCS

scientific retrieval en

Citation prediction, document classification, and recommendation for scientific papers

Corpus: 25,656 Queries: 1,000
Quality
map at 10 0.1065
mrr at 10 0.3257
ndcg at 10 0.1833
Reference →

SCIDOCS (candidates: gte-multilingual-base, k=50)

scientific retrieval en

Citation prediction, document classification, and recommendation for scientific papers

Performance L4 b1 c16
Query 53.7K tok/s
Query p50 42.5ms
Reference →

SciFact

scientific retrieval en

Scientific claim verification using research literature

Corpus: 5,183 Queries: 300
Quality
map at 10 0.6569
mrr at 10 0.6764
ndcg at 10 0.7071
Reference →

SciFact (candidates: gte-multilingual-base, k=50)

scientific retrieval en

Scientific claim verification using research literature

Performance L4 b1 c16
Query 67.4K tok/s
Query p50 42.1ms
Reference →

StackOverflowQA

technology retrieval en

Programming question answering from Stack Overflow

Corpus: 19,931 Queries: 1,994
Quality
map at 10 0.5244
mrr at 10 0.5349
ndcg at 10 0.5815
Reference →

StackOverflowQA (candidates: gte-multilingual-base, k=50)

technology retrieval en

Programming question answering from Stack Overflow

Performance L4 b1 c16
Query 98.6K tok/s
Query p50 47.2ms
Reference →

T2Reranking

general reranking zh

Chinese passage ranking benchmark

Quality
map at 10 0.4714
mrr at 10 0.7102
Reference →

Open source inference for agents

Open-source inference for the models behind your agents. Run it yourself, or let us run it for you.

Github 3.3K

Contact us

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

Apply for an inference grant

Free capacity on our hosted cluster for selected projects. Tell us what you run and we reply by email.