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

← Catalog

openai/clip-vit-base-patch32

Open comparison →

Primitive: /encode · Encode · CLIP

Disclaimer: The model card is taken and modified from the official CLIP repository, it can be found here.

MultimodalDense

Overview

Hardware: — drives latency, throughput & cost

Size151M params
Tasks /encode
License
Latency234 ms
Throughput958 tok/s
Cost$0.232 /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: 512
Max sequence length77
Inputstext · image

Benchmarks

Flickr30kI2TRetrieval

general retrieval en

Image-to-text retrieval: retrieve captions from images

Corpus: 31,783 Queries: 1,000
Quality
ndcg at 10 0.7165
map at 10 0.6029
mrr at 10 0.8521
Performance L4 b1 c16
Corpus 958 tok/s
Corpus p50 234.0ms
Query 10.0 mpix/s
Query p50 245.6ms
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.