S

Blip Image Captioning Base

Aug 2023
SBlip Image Captioning Base
Compare in
on
self-host onlyprice
1.1MIEB lite
Aug 2023released
768Dimensions
1K tokensMax input

Ranking?

Full ranking →
1LCO Embedding Omni 7B502LCO Embedding Omni 3B47.63BidirLM Omni 2.5B Embedding464Ebind Full415Jina Embeddings v5 Omni Small40.5
···
44Blip2 Opt 2.7B2445Blip Image Captioning Base23.746Bge Visualized Base23.3

Strong at

C
Embeddings & Retrieval#89 of 89

Alternatives?

ModelScoreCostPriceWhat changes
BetterEmbeddingGemma 2unstatedOpen+63%—$0.02Memory (4-bit): 0.1 GB → 0.2 GBLOpenLCO Embedding Omni 7Bunstated×2.1——Memory (4-bit): 0.1 GB → 5.4 GBBetterJina Embeddings v5 Omni NanounstatedOpen+25%—$0.02Memory (4-bit): 0.1 GB → 0.6 GBBetterJina Embeddings v5 Omni SmallunstatedOpen+71%—$0.05Memory (4-bit): 0.1 GB → 1 GBBetterJina Clip v1Open+65%—$0.05

Run it yourself

estimate
Fits on✓ 8 GB (phone, laptop) · full quality✓ 16 GB · full quality✓ 24 GB GPU · full quality✓ 48 GB GPU · full quality✓ 80 GB GPU · full quality
LicenceBSD-3-Clausefree for commercial use

Benchmarks

in Embeddings & Retrieval
MIEB English · Document Understanding · English0.9#80
MIEB English · Zero-shot Classification · English9.4#38
MIEB English · Visual semantic similarity (English) · multilingual (13 languages)55.4#36
MIEB English · Image Classification · English58.2#45
MIEB English · Compositionality · English27.6#38
MIEB English · Vision Centric QA · English43.4#37
MIEB English · Image Clustering · English63.1#44
MIEB Img · Image Classification · English58.2#45
MIEB Img · Visual semantic similarity (English) · English53.1#36
MIEB Img · Visual semantic similarity (multilingual) · multilingual (13 languages)34.9#33
MIEB Img · Image Clustering · English63.1#44
MIEB lite · Mean over tasks · multilingual (41 languages)27.6#47
MIEB lite · Mean over task types · multilingual (41 languages)27.6#47
MIEB lite · Vision Centric QA · English45.6#43
MIEB lite · Any-to-any Retrieval · English10.7#47
MIEB lite · Document Understanding · English1.1#89
MIEB lite · Zero-shot Classification · English3.4#47
MIEB lite · Compositionality · English23.7#45
MIEB lite · Image Classification · English53#53
MIEB lite · Any-to-any Multilingual Retrieval · multilingual (40 languages)0.3#48
MIEB lite · Image Clustering · English53.3#50
MIEB lite · Visual semantic similarity (English) · English53.7#45
MIEB lite · Visual semantic similarity (multilingual) · multilingual (13 languages)31.2#37
MIEB multilingual · Zero-shot Classification · English9.4#38
MIEB multilingual · Image Classification · English58.2#45
MIEB multilingual · Document Understanding · English0.9#80
MIEB multilingual · Compositionality · English27.6#38
MIEB multilingual · Vision Centric QA · English43.4#37
MIEB multilingual · Visual semantic similarity (English) · multilingual (13 languages)55.4#36
MIEB multilingual · Visual semantic similarity (multilingual) · multilingual (13 languages)31.2#37
MIEB multilingual · Any-to-any Multilingual Retrieval · multilingual (40 languages)0.3#39
MIEB multilingual · Image Clustering · English63.1#44

FAQ

How much does Blip Image Captioning Base cost?›

There is no per-unit API price: self-host only.

Is Blip Image Captioning Base free to use?›

There is no free API. The weights are open, so you can run it yourself.

How good is Blip Image Captioning Base, and what is it best at?›

#89 of 89 in Multimodal.

Is Blip Image Captioning Base better than Blip2 Opt 2.7B?›

On MIEB lite · Compositionality · English, Blip Image Captioning Base scores 23.7 vs 24 — about the same.

When was Blip Image Captioning Base released? Is there a newer version?›

Released 2023-08-01. It is the current version in its line.

Is Blip Image Captioning Base open source?›

Yes, the weights are open (bsd-3-clause licence).

Can I run Blip Image Captioning Base locally?›

The weights are open and can be self-hosted; hardware needs are not confirmed yet.

What is Blip Image Captioning Base?›

Blip Image Captioning Base is an open-weights embeddings & retrieval model from Salesforce, released Aug 2023.

What are the best alternatives to Blip Image Captioning Base?›

EmbeddingGemma 2 (better), LCO Embedding Omni 7B (open weights).