Benchmark By Score Rank
MIEB liteDocument Understanding · English Embeddings match queries to document content on MIEB Lite · Document Understanding in English MTEB 28.7 #68 of 89+11 versions & slices Visual semantic similarity (multilingual) · multilingual (13 languages) Embeddings estimate visual semantic similarity on MIEB Lite · Visual STS (multilingual) in Arabic, Dutch, English, French, German, Italian, Korean, Mandarin Chinese, Polish, Portuguese, Russian, Spanish, Turkish MTEB 26 #44 of 56Visual semantic similarity (English) · English Embeddings estimate visual semantic similarity on MIEB Lite · Visual STS (English) in English MTEB 68.7 #24 of 54Image Classification · English Embeddings classify images on MIEB Lite · Image Classification in English MTEB 75.9 #8 of 53Image Clustering · English Embeddings group similar images on MIEB Lite · Image Clustering in English MTEB 76.4 #18 of 53Any-to-any Multilingual Retrieval · multilingual (40 languages) Embeddings retrieve matching examples across languages and modalities on MIEB Lite · Any-to-any Multilingual Retrieval in 40 languages including Arabic, Bengali, Bulgarian, Chinese, Croatian, Cusco Quechua, Czech, Danish MTEB 21.4 #30 of 48Vision Centric QA · English Embeddings match visual content to questions on MIEB Lite · Vision Centric QA in English MTEB 46.9 #38 of 48Mean over tasks · multilingual (41 languages) Embeddings match visual content to questions, retrieve matching examples across modalities, match queries to document content, assign unseen category labels to examples, distinguish changes in how objects and attributes are combined, classify images, retrieve matching examples across languages and modalities, group similar images, estimate visual semantic similarity, estimate visual semantic similarity on MIEB Lite · mean over tasks in 41 languages including Arabic, Bengali, Bulgarian, Chinese, Croatian, Cusco Quechua, Czech, Danish; tasks receive equal weight MTEB 47.6 #26 of 47Mean over task types · multilingual (41 languages) Embeddings match visual content to questions, retrieve matching examples across modalities, match queries to document content, assign unseen category labels to examples, distinguish changes in how objects and attributes are combined, classify images, retrieve matching examples across languages and modalities, group similar images, estimate visual semantic similarity, estimate visual semantic similarity on MIEB Lite · mean over task types in 41 languages including Arabic, Bengali, Bulgarian, Chinese, Croatian, Cusco Quechua, Czech, Danish; task types receive equal weight MTEB 47.6 #26 of 47Any-to-any Retrieval · English Embeddings retrieve matching examples across modalities on MIEB Lite · Any-to-any Retrieval in English MTEB 31.1 #16 of 47Compositionality · English Embeddings distinguish changes in how objects and attributes are combined on MIEB Lite · Compositionality in English MTEB 33.6 #23 of 47Zero-shot Classification · English Embeddings assign unseen category labels to examples on MIEB Lite · Zero-shot Classification in English MTEB 67.8 #11 of 47
MIEB EnglishDocument Understanding · English Embeddings match queries to document content on MIEB English · Document Understanding in English MTEB 36.3 #60 of 81+9 versions & slices Visual semantic similarity (English) · multilingual (13 languages) Embeddings estimate visual semantic similarity on MIEB English · Visual STS (English) in Arabic, Dutch, English, French, German, Italian, Korean, Mandarin Chinese, Polish, Portuguese, Russian, Spanish, Turkish MTEB 65.8 #22 of 46Image Classification · English Embeddings classify images on MIEB English · Image Classification in English MTEB 81.2 #9 of 45Image Clustering · English Embeddings group similar images on MIEB English · Image Clustering in English MTEB 83.5 #7 of 45Vision Centric QA · English Embeddings match visual content to questions on MIEB English · Vision Centric QA in English MTEB 45.9 #28 of 40Compositionality · English Embeddings distinguish changes in how objects and attributes are combined on MIEB English · Compositionality in English MTEB 40.8 #15 of 39Zero-shot Classification · English Embeddings assign unseen category labels to examples on MIEB English · Zero-shot Classification in English MTEB 65.8 #9 of 39Mean over tasks · multilingual (13 languages) Embeddings match queries to document content, assign unseen category labels to examples, retrieve matching examples across modalities, estimate visual semantic similarity, classify images, distinguish changes in how objects and attributes are combined, match visual content to questions, group similar images on MIEB English · mean over tasks in Arabic, Dutch, English, French, German, Italian, Korean, Mandarin Chinese, Polish, Portuguese, Russian, Spanish, Turkish; tasks receive equal weight MTEB 57.2 #15 of 37Mean over task types · multilingual (13 languages) Embeddings match queries to document content, assign unseen category labels to examples, retrieve matching examples across modalities, estimate visual semantic similarity, classify images, distinguish changes in how objects and attributes are combined, match visual content to questions, group similar images on MIEB English · mean over task types in Arabic, Dutch, English, French, German, Italian, Korean, Mandarin Chinese, Polish, Portuguese, Russian, Spanish, Turkish; task types receive equal weight MTEB 57.2 #15 of 37Any-to-any Retrieval · English Embeddings retrieve matching examples across modalities on MIEB English · Any-to-any Retrieval in English MTEB 38 #12 of 37
MIEB multilingualDocument Understanding · English Embeddings match queries to document content on MIEB Multilingual · Document Understanding in English MTEB 36.3 #60 of 81+11 versions & slices Visual semantic similarity (multilingual) · multilingual (13 languages) Embeddings estimate visual semantic similarity on MIEB Multilingual · Visual STS (multilingual) in Arabic, Dutch, English, French, German, Italian, Korean, Mandarin Chinese, Polish, Portuguese, Russian, Spanish, Turkish MTEB 26 #44 of 56Visual semantic similarity (English) · multilingual (13 languages) Embeddings estimate visual semantic similarity on MIEB Multilingual · Visual STS (English) in Arabic, Dutch, English, French, German, Italian, Korean, Mandarin Chinese, Polish, Portuguese, Russian, Spanish, Turkish MTEB 65.8 #22 of 46Image Classification · English Embeddings classify images on MIEB Multilingual · Image Classification in English MTEB 81.2 #9 of 45Image Clustering · English Embeddings group similar images on MIEB Multilingual · Image Clustering in English MTEB 83.5 #7 of 45Vision Centric QA · English Embeddings match visual content to questions on MIEB Multilingual · Vision Centric QA in English MTEB 45.9 #28 of 40Any-to-any Multilingual Retrieval · multilingual (40 languages) Embeddings retrieve matching examples across languages and modalities on MIEB Multilingual · Any-to-any Multilingual Retrieval in 40 languages including Arabic, Bengali, Bulgarian, Chinese, Croatian, Cusco Quechua, Czech, Danish MTEB 23 #21 of 39Compositionality · English Embeddings distinguish changes in how objects and attributes are combined on MIEB Multilingual · Compositionality in English MTEB 40.8 #15 of 39Zero-shot Classification · English Embeddings assign unseen category labels to examples on MIEB Multilingual · Zero-shot Classification in English MTEB 65.8 #9 of 39Mean over tasks · multilingual (41 languages) Embeddings assign unseen category labels to examples, classify images, match queries to document content, retrieve matching examples across modalities, distinguish changes in how objects and attributes are combined, match visual content to questions, estimate visual semantic similarity, estimate visual semantic similarity, retrieve matching examples across languages and modalities, group similar images on MIEB Multilingual · mean over tasks in 41 languages including Arabic, Bengali, Bulgarian, Chinese, Croatian, Cusco Quechua, Czech, Danish; tasks receive equal weight MTEB 50.6 #19 of 37Mean over task types · multilingual (41 languages) Embeddings assign unseen category labels to examples, classify images, match queries to document content, retrieve matching examples across modalities, distinguish changes in how objects and attributes are combined, match visual content to questions, estimate visual semantic similarity, estimate visual semantic similarity, retrieve matching examples across languages and modalities, group similar images on MIEB Multilingual · mean over task types in 41 languages including Arabic, Bengali, Bulgarian, Chinese, Croatian, Cusco Quechua, Czech, Danish; task types receive equal weight MTEB 50.6 #19 of 37Any-to-any Retrieval · English Embeddings retrieve matching examples across modalities on MIEB Multilingual · Any-to-any Retrieval in English MTEB 38 #12 of 37
MIEB ImgVisual semantic similarity (multilingual) · multilingual (13 languages) Embeddings estimate visual semantic similarity on MIEB Img · Visual STS (multilingual) in Arabic, Dutch, English, French, German, Italian, Korean, Mandarin Chinese, Polish, Portuguese, Russian, Spanish, Turkish MTEB 31.4 #37 of 50+6 versions & slices Visual semantic similarity (English) · English Embeddings estimate visual semantic similarity on MIEB Img · Visual STS (English) in English MTEB 65.1 #21 of 46Image Classification · English Embeddings classify images on MIEB Img · Image Classification in English MTEB 81.2 #9 of 45Image Clustering · English Embeddings group similar images on MIEB Img · Image Clustering in English MTEB 83.5 #7 of 45Any-to-any Retrieval · English Embeddings retrieve matching examples across modalities on MIEB Img · Any-to-any Retrieval in English MTEB 38.5 #13 of 42Mean over tasks · multilingual (13 languages) Embeddings classify images, retrieve matching examples across modalities, estimate visual semantic similarity, estimate visual semantic similarity, group similar images on MIEB Img · mean over tasks in Arabic, Dutch, English, French, German, Italian, Korean, Mandarin Chinese, Polish, Portuguese, Russian, Spanish, Turkish; tasks receive equal weight MTEB 59.9 #14 of 40Mean over task types · multilingual (13 languages) Embeddings classify images, retrieve matching examples across modalities, estimate visual semantic similarity, estimate visual semantic similarity, group similar images on MIEB Img · mean over task types in Arabic, Dutch, English, French, German, Italian, Korean, Mandarin Chinese, Polish, Portuguese, Russian, Spanish, Turkish; task types receive equal weight MTEB 59.9 #14 of 40
No benchmark matches
BetterEmbeddingGemma 2unstatedOpen×3—$0.02LOpenLCO Embedding Omni 7Bunstated×3.2——Memory (4-bit): 0.3 GB → 5.4 GB
BetterJina Embeddings v5 Omni NanounstatedOpen×2.1—$0.02Memory (4-bit): 0.3 GB → 0.6 GB