Skip to content

Benchmark

MTEB(Multilingual, v1) - Retrieval

Embedding benchmarks measure how well models find related items, for example when matching a search to relevant results. Scores are percentages.

Models measured
35
Scores recorded
35
Sources
1
IQ weight
0.73

Score distribution

52%
Mean
55.8%
Median
14%
Std dev
35.9%
10th percentile
67.7%
90th percentile
10.7%
Min
10.7%70.9%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.73 for embedding models, measured over 35 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Qwen3 Embedding 8BRetrieval· MTEB Leaderboard data70.9%
2Embed v1 4bRetrieval· MTEB Leaderboard data69.7%
3Qwen3 Embedding 4BRetrieval· MTEB Leaderboard data69.6%
4Gemini Embedding 001Retrieval· MTEB Leaderboard data67.7%
5INF Retriever v1Retrieval· MTEB Leaderboard data66.5%
6Embed v1 0.6bRetrieval· MTEB Leaderboard data65.4%
7Jina Embeddings v5 Omni SmallRetrieval· MTEB Leaderboard data64.9%
8Jina Embeddings v5 Text SmallRetrieval· MTEB Leaderboard data64.9%
9Qwen3 Embedding 0.6BRetrieval· MTEB Leaderboard data64.7%
10Jina Embeddings v5 Omni NanoRetrieval· MTEB Leaderboard data63.3%
11Jina Embeddings v5 Text NanoRetrieval· MTEB Leaderboard data63.3%
12embeddinggemma-2Retrieval· MTEB Leaderboard data60.6%
13Text Multilingual Embedding 002Retrieval· MTEB Leaderboard data59.7%
14Text Embedding 3 LargeRetrieval· MTEB Leaderboard data59.3%
15snowflake-arctic-embed-l-v2.0Retrieval· MTEB Leaderboard data58.4%
16Multilingual E5 Large InstructRetrieval· MTEB Leaderboard data57.1%
17NV-Embed-v2Retrieval· MTEB Leaderboard data56.7%
18Jina Embeddings v3Retrieval· MTEB Leaderboard data55.8%
19E5 Mistral 7B InstructRetrieval· MTEB Leaderboard data55.8%
20snowflake-arctic-embed-m-v2.0Retrieval· MTEB Leaderboard data54.8%
21BGE M3Retrieval· MTEB Leaderboard data54.6%
22Granite Embedding 278m MultilingualRetrieval· MTEB Leaderboard data52.2%
23UAE-Large-V1Retrieval· MTEB Leaderboard data41.2%
24E5 Base V2Retrieval· MTEB Leaderboard data40.5%
25E5 Large V2Retrieval· MTEB Leaderboard data40.1%
26Paraphrase Multilingual MPNet v2Retrieval· MTEB Leaderboard data39.8%
27BGE Large En V1.5Retrieval· MTEB Leaderboard data39%
28BGE Base En V1.5Retrieval· MTEB Leaderboard data38.6%
29Gte LargeRetrieval· MTEB Leaderboard data38.2%
30nomic-embed-text-v1Retrieval· MTEB Leaderboard data37.1%
31bge-small-en-v1.5Retrieval· MTEB Leaderboard data36.3%
32Gte BaseRetrieval· MTEB Leaderboard data35.9%
33nomic-embed-text-v1.5Retrieval· MTEB Leaderboard data34.1%
34All MiniLM L6 V2Retrieval· MTEB Leaderboard data32.5%
35Jina Embeddings v2 Base ENRetrieval· MTEB Leaderboard data10.7%