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

← Catalog

Alibaba-NLP/gte-Qwen2-7B-instruct

Open comparison →

Primitive: /encode · Encode · Qwen2

gte-Qwen2-7B-instruct is the latest model in the gte (General Text Embedding) model family that ranks No.1 in both English and Chinese evaluations on the Massive Text Embedding Benchmark MTEB benchmark (as of June 16, 2024).

Long contextDense

Overview

Hardware: — drives latency, throughput & cost

Size7.6B params
Tasks /encode
Licenseapache-2.0
Latency846 ms
Throughput3.5K tok/s
Cost$0.063 /1M tok

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

Embedding

Output typesDense
Dimensionsdense: 3,584
Max sequence length32,000
Inputstext

Benchmarks

NFCorpus

medical retrieval en

Biomedical literature search from NutritionFacts.org

Corpus: 3,593 Queries: 323
Quality
ndcg at 10 0.4040
map at 10 0.1548
mrr at 10 0.6133
Performance L4 b1 c16
Corpus 3.7K tok/s
Corpus p50 1.1s
Query 228 tok/s
Query p50 361.2ms
Reference →

NanoFiQA2018Retrieval

finance retrieval en

Smaller subset of the FiQA financial QA dataset

Quality
ndcg at 10 0.6902
map at 10 0.6156
mrr at 10 0.7338
Performance L4 b1 c16
Corpus 3.3K tok/s
Corpus p50 594.5ms
Query 500 tok/s
Query p50 221.7ms
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 2.3K

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.