Skip to content

Benchmark

MTEB(pol, v1) - Mean (TaskType)

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

Models measured
17
Scores recorded
17
Sources
1
IQ weight
1.18

Score distribution

56.3%
Mean
51.5%
Median
9.2%
Std dev
47.7%
10th percentile
68.7%
90th percentile
44.2%
Min
44.2%71.7%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.18 for embedding models, measured over 17 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 8BMean (TaskType)· MTEB Leaderboard data71.7%
2Qwen3 Embedding 4BMean (TaskType)· MTEB Leaderboard data70.5%
3E5 Mistral 7B InstructMean (TaskType)· MTEB Leaderboard data68.7%
4Multilingual E5 Large InstructMean (TaskType)· MTEB Leaderboard data68.4%
5Qwen3 Embedding 0.6BMean (TaskType)· MTEB Leaderboard data64.6%
6snowflake-arctic-embed-l-v2.0Mean (TaskType)· MTEB Leaderboard data61%
7Paraphrase Multilingual MPNet v2Mean (TaskType)· MTEB Leaderboard data61%
8Granite Embedding 278m MultilingualMean (TaskType)· MTEB Leaderboard data55.8%
9E5 Large V2Mean (TaskType)· MTEB Leaderboard data51.5%
10E5 Base V2Mean (TaskType)· MTEB Leaderboard data49.6%
11Gte LargeMean (TaskType)· MTEB Leaderboard data49.5%
12UAE-Large-V1Mean (TaskType)· MTEB Leaderboard data49.3%
13Gte BaseMean (TaskType)· MTEB Leaderboard data48.6%
14BGE Large En V1.5Mean (TaskType)· MTEB Leaderboard data48.2%
15BGE Base En V1.5Mean (TaskType)· MTEB Leaderboard data47.7%
16bge-small-en-v1.5Mean (TaskType)· MTEB Leaderboard data47.6%
17All MiniLM L6 V2Mean (TaskType)· MTEB Leaderboard data44.2%