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

← Catalog

BAAI/bge-m3

Open comparison →

Primitive: /score · Score · XLM-RoBERTa

For more details please refer to our github repo: https://github.com/FlagOpen/FlagEmbedding

Long context

Overview

Hardware: — drives latency, throughput & cost

Size568M params
Tasks /encode · /score
Licensemit
Latency56 ms
Throughput2.9K tok/s
Cost$0.076 /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 length8,192

Benchmarks

AskUbuntuDupQuestions

technology reranking en

Duplicate question detection from AskUbuntu

Corpus: 6,743 Queries: 360
Quality
ndcg at 10 0.6657
map at 10 0.5141
mrr at 10 0.7429
Performance L4 b1 c16
Corpus 3.3K tok/s
Corpus p50 29.2ms
Query 3.2K tok/s
Query p50 29.3ms
Reference →

CQADupstackPhysicsRetrieval

scientific retrieval en

Duplicate question retrieval from StackExchange Physics

Corpus: 38,314 Queries: 1,039
Performance L4 b1 c16
Corpus 27.0K tok/s
Corpus p50 75.3ms
Query 2.6K tok/s
Query p50 55.2ms
Reference →

CosQA

technology retrieval en

Code search with natural language queries

Corpus: 6,267 Queries: 500
Performance L4 b1 c16
Corpus 13.5K tok/s
Corpus p50 58.4ms
Query 1.5K tok/s
Query p50 57.6ms
Reference →

FiQA2018

finance retrieval en

Financial opinion mining and question answering

Corpus: 57,599 Queries: 648
Performance L4 b1 c16
Corpus 26.8K tok/s
Corpus p50 89.3ms
Query 2.9K tok/s
Query p50 55.8ms
Reference →

LegalBenchConsumerContractsQA

legal retrieval en

Question answering on consumer contracts

Corpus: 153 Queries: 396
Performance L4 b1 c16
Corpus 36.7K tok/s
Corpus p50 224.8ms
Query 3.7K tok/s
Query p50 59.5ms
Reference →

MMarcoReranking

general reranking zh

Multilingual MARCO passage reranking (Chinese)

Quality
ndcg at 10 0.3153
map at 10 0.2609
mrr at 10 0.2620
Performance L4 b1 c16
Corpus 43.5K tok/s
Corpus p50 30.1ms
Query 3.5K tok/s
Query p50 29.5ms
Reference →

NFCorpus

medical retrieval en

Biomedical literature search from NutritionFacts.org

Corpus: 3,593 Queries: 323
Performance L4 b1 c16
Corpus 34.4K tok/s
Corpus p50 137.4ms
Query 1.2K tok/s
Query p50 56.8ms
Reference →

NanoFiQA2018Retrieval

finance retrieval en

Smaller subset of the FiQA financial QA dataset

Performance L4 b1 c16
Corpus 30.9K tok/s
Corpus p50 86.4ms
Query 2.8K tok/s
Query p50 54.9ms
Reference →

SCIDOCS

scientific retrieval en

Citation prediction, document classification, and recommendation for scientific papers

Corpus: 25,656 Queries: 1,000
Performance L4 b1 c16
Corpus 31.0K tok/s
Corpus p50 101.1ms
Query 2.7K tok/s
Query p50 55.7ms
Reference →

SciFact

scientific retrieval en

Scientific claim verification using research literature

Corpus: 5,183 Queries: 300
Performance L4 b1 c16
Corpus 32.2K tok/s
Corpus p50 126.9ms
Query 3.9K tok/s
Query p50 57.5ms
Reference →

StackOverflowQA

technology retrieval en

Programming question answering from Stack Overflow

Corpus: 19,931 Queries: 1,994
Performance L4 b1 c16
Corpus 29.5K tok/s
Corpus p50 118.6ms
Query 31.5K tok/s
Query p50 147.9ms
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 2.1K

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.