ModelCards
← All benchmarks

MTEB English v2, mean (task)

We haven't written a description for this benchmark yet; see the card notes below for how each maker ran it.

Your workloadRAG over docs · 12K in · 600 out · 50K requests/mo · 30% cached
  1. 1Qwen3-Embedding-8B · Alibaba
    75.22 score
  2. 2EmbeddingGemma 2 · Google
    68.46 score

“Best value”: no other model here reports a higher score at a lower cost for your workload. Scores are self-reported by each maker with different settings, so treat close calls as ties.

How each maker reported it

  • Qwen3-Embedding-8B technical report Tables 2–3; model card

    Alibaba's own numbers. Competitor scores were taken from the MTEB leaderboard in early June 2025. CMTEB for the 8B is 73.84 in the model card and Table 8 vs 73.83 in Table 3.

  • EmbeddingGemma 2 model card (768-dim, full-precision checkpoint)

    Google's own numbers. MTEB English v2 (768d): 68.46. MRL truncation keeps quality close to lossless down to 256 dims; 128 dims degrades multimodal quality substantially (e.g. MMEB v2 overall 59.01 at 768d vs 45.65 at 128d).

Only each model's own maker's number is shown; scores a maker reports for rival models are left out.