Benchmark Almanac

Qwen3-Embedding-8B

Qwen3-Embedding-8B ranks first on 7 of the 14 benchmarks with five or more models scored.

47
Benchmarks with a score
7
Ranked first, out of five or more models
12
In the top three
Qwen Jun 2025

Qwen3-Embedding-8B

Open weights

No Epoch Capabilities Index score yet
  • CLICK-ID 65.1% #1 /5
  • Estonian Valence Corpus 59.9% #1 /5
  • Greek Legal Code 56.5% #1 /5

Where has it been number one?

Qwen3-Embedding-8B is number one on 1 benchmark today. It also led 2 more before newer models passed it. It still tops 3 that nobody has tested since.

Where it ranks

Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.

Knowledge and language

BenchmarkScoreRankSource
Ask Ubuntu duplicate question retrieval 0.70 #1 of 5 Official leaderboard
CLICK-ID 65.1% #1 of 5 Official leaderboard
Estonian Valence Corpus 59.9% #1 of 5 Official leaderboard
Amazon Counterfactual Detection 93.2% #1 of 4 Official leaderboard
Corpus for Emotion Detection in Russian 53% #1 of 4 Official leaderboard
CTKFacts 0.89 #1 of 4 Official leaderboard
Faroese Semantic Textual Similarity 0.87 #1 of 4 Official leaderboard
IndicLID benchmark 95.1% #1 of 4 Official leaderboard
JSICK 89.4% #1 of 4 Official leaderboard
Korean Sarcasm Dataset 62.1% #1 of 4 Official leaderboard
MTEB-BR 0.67 #1 of 4 Official leaderboard
ScaLA 57.4% #1 of 4 Official leaderboard
SemEval 2022 Task 8 multilingual news article similarity 0.72 #1 of 4 Official leaderboard
BRIGHT-Pro 41.8 #1 of 3 Official leaderboard
FarSick 0.76 #1 of 3 Official leaderboard
Q2D-Web 79.3% #1 of 3 Official leaderboard
DBpedia 14 99.3% #2 of 5 Official leaderboard
Tatoeba sentence retrieval 79.7% #2 of 5 Official leaderboard
FollowIR 0.10 #2 of 4 Official leaderboard
SemEval 2024 Task 1 semantic textual relatedness 0.64 #2 of 4 Official leaderboard
DiaBLa 0.87 #3 of 5 Official leaderboard
IN22-Gen 82% #3 of 5 Official leaderboard
Norwegian Courts Bitext Mining 93% #4 of 6 Official leaderboard
CLIMATE-FEVER 0.47 #2 of 3 Official leaderboard
NaijaSenti 48.8% #1 of 1 Official leaderboard

Math and science

BenchmarkScoreRankSource
ScholarCatalyst 37% #4 of 5 Official leaderboard

Coding

BenchmarkScoreRankSource
CodeSearchNet Challenge 0.93 #1 of 4 Official leaderboard
CoSQA 0.38 #2 of 4 Official leaderboard
CodeTransOcean 32.8 #3 of 4 Official leaderboard

Professional work

BenchmarkScoreRankSource
AILA Statute Retrieval 0.85 #1 of 5 Official leaderboard
BEIR FiQA-2018 Retrieval 0.65 #1 of 5 Official leaderboard
Financial PhraseBank 95.2% #1 of 5 Official leaderboard
Greek Legal Code 56.5% #1 of 5 Official leaderboard
Bank Question Corpus 0.76 #1 of 4 Official leaderboard
GerDaLIR Small 0.45 #1 of 4 Official leaderboard
GerLayQA 0.27 #1 of 4 Official leaderboard
GerLeRB 0.70 #1 of 4 Official leaderboard
LegalQuAD 0.70 #1 of 4 Official leaderboard
NFCorpus 0.41 #1 of 4 Official leaderboard
OpenLegalRetrieval-DE 0.44 #1 of 4 Official leaderboard
BANKING77 87.3% #2 of 5 Official leaderboard
GreekBarRetrieval 0.28 #2 of 4 Official leaderboard
MultiEURLEX 0.05 #2 of 4 Official leaderboard
BuiltBench 65.3 #1 of 2 Official leaderboard
Italian-Administrative-CaseHold 66.1% #3 of 4 Official leaderboard

Other

BenchmarkScoreRankSource
MTEB 72.2% #1 of 4 Official leaderboard
RTEB 69.4% #1 of 2 Official leaderboard

Inside agents

Boards that list Qwen3-Embedding-8B inside an agent or product, such as Reform- Qwen3-8B (Qwen3-Embedding-8B + gpt-oss-120b). Those scores belong to the agent, so we don't count them as the model's own. Its own score on the same board is beside each one.

BenchmarkEntryAgent's scoreIts own score
GreekBarRetrieval Reform- Qwen3-8B (Qwen3-Embedding-8B + gpt-oss-120b) 0.33 0.28

Who measured it?

Independent runs cover all 47 of its benchmarks.

  • An independent run only 47
Official leaderboards scored it on 47.

Sources: Official leaderboards, Epoch AI (CC BY 4.0) and labs’ model cards