Benchmarks · /benchmarks/mteb-retrieval-en-v2

Retrieval

MTEB retrieval slice for embeddings.

Source · MTEB
Version · mteb snapshot 2026-05-13
Scores · 11

Passport

Visible tradeoffsThis is a retrieval signal, so it is best read as search-stack quality rather than broad model capability.

source

MTEB

metric

NDCG@10 (ndcg)

judge

Retrieval

direction

higher better

group id

mteb_retrieval_en_v2

domain

Embeddings / retrieval

What it measures vs what it misses

✓ Measures

Embedding quality for retrieval tasks.

✗ Misses

Chat quality, generation, latency.

Why this countsIt is one of the few direct signals for retrieval stacks, where embedding quality matters more than chat style.Comparable-group ruleThis percentile only compares models inside the exact benchmark/version group shown here. It is not a universal score.What it missesIt does not tell you whether the same model is strong at generation, ranking policy, or final answer quality.

Leaderboard · this benchmark version

#1 · GPT-5

MTEB · Mar 19, 2026

58.8 ndcg

#2 · GPT-5.4

MTEB · Mar 19, 2026

58.8 ndcg

#3 · GPT-5.4 mini

MTEB · Mar 19, 2026

58.8 ndcg

#4 · GPT-5.4 nano

MTEB · Mar 19, 2026

58.8 ndcg

#5 · Gemini 2.5 Pro

MTEB · Mar 19, 2026

57.9 ndcg

#6 · Qwen3 235B A22B

MTEB · Mar 19, 2026

56.2 ndcg

#7 · Grok 4

MTEB · Mar 19, 2026

52.1 ndcg

#8 · Mistral Medium 3

MTEB · Mar 19, 2026

51.3 ndcg

#9 · Llama 4 Maverick

MTEB · Mar 19, 2026

49.6 ndcg

#10 · BAAI bge-large-en-v1.5

MTEB · May 13, 2026

49.3 ndcg

#11 · deepseek-r1-0528

MTEB · Mar 19, 2026

48.4 ndcg