Q2D-Web
Retrievers search 190 million web documents for queries that agents reformulate from real conversations.
The board measures recall at 1,000 against three sets of relevance judgments across ten languages.
- Made by
- Perplexity
- Released
- September 2026
- Kind
- Questions that need search
- Level
- University
- Scored
- Checked automatically
- Answers
- Prediction
- Tasks
- Kept private
- Measures
- Capability
- Language
- Many languages
- Interaction
- Single turn
- Items from
- Real work and use, Generated
- Refresh
- One fixed set
- Run by
- A host
- In views
- Many languages
- Last model added
- June 2025
- Scores refreshed
- November 20th, 2026
perplexity-ai/pplx-embed-v2-late-9b tops the leaderboard with 88.2% Qwen3-Embedding-8B leads with 79.3% among models released by June 2025
Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. The best model we track is Qwen3-Embedding-8B, at 79.3%. This view shows only models we track, each at its best.
The newest model with a score here came out in June 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 88.2 perplexity-ai/pplx-embed-v2-late-9b
- 86.7 perplexity-ai/pplx-embed-v2-late-0.6b
- 85.0 perplexity-ai/pplx-embed-v1-4b
- 83.1 nvidia/Nemotron-3-Embed-8B-BF16
- 81.8 perplexity-ai/pplx-embed-v1-0.6b
- 80.1 google/embeddinggemma-300m
- 79.3
Qwen3-Embedding-8B
- 79.2 topk-io/topk-embed-v1-small (128 dim output)
- 78.6 perplexity-ai/pplx-embed-v1-late-0.6b
- 77.4 topk-io/topk-embed-v1-xsmall (128 dim output)
- 76.5
Qwen3-Embedding-4B
- 75.6 nvidia/Nemotron-3-Embed-1B-BF16
- 74.8 jinaai/jina-embeddings-v5-text-small
- 74.2 lightonai/mLateOn
- 73.7 lightonai/mDenseOn
- 72.6 chungimungi/GLInt
- 71.2
Qwen3-Embedding-0.6B
- 69.5 voyageai/voyage-4-nano
- 60.8 BM25-tantivy
- 1 perplexity-ai/pplx-embed-v2-late-9b 88.2%
- 2 perplexity-ai/pplx-embed-v2-late-0.6b 86.7%
- 3 perplexity-ai/pplx-embed-v1-4b 85%
- 4 nvidia/Nemotron-3-Embed-8B-BF16 83.1%
- 5 perplexity-ai/pplx-embed-v1-0.6b 81.8%
- 6 google/embeddinggemma-300m 80.1%
- 7
Qwen3-Embedding-8B 79.3%
- 8 topk-io/topk-embed-v1-small (128 dim output) 79.2%
- 9 perplexity-ai/pplx-embed-v1-late-0.6b 78.6%
- 10 topk-io/topk-embed-v1-xsmall (128 dim output) 77.4%
- 11
Qwen3-Embedding-4B 76.5%
- 12 nvidia/Nemotron-3-Embed-1B-BF16 75.6%
- 13 jinaai/jina-embeddings-v5-text-small 74.8%
- 14 lightonai/mLateOn 74.2%
- 15 lightonai/mDenseOn 73.7%
- 16 chungimungi/GLInt 72.6%
- 17
Qwen3-Embedding-0.6B 71.2%
- 18 voyageai/voyage-4-nano 69.5%
- 19 BM25-tantivy 60.8%
Source: Official Q2D-Web leaderboard
- 1
Qwen3-Embedding-8B 79.3%
- 2
Qwen3-Embedding-4B 76.5%
- 3
Qwen3-Embedding-0.6B 71.2%
Source: Official Q2D-Web leaderboard
| # | Entry | Score | Setting | Date | Link | Source |
|---|---|---|---|---|---|---|
| 1 | perplexity-ai/pplx-embed-v2-late-9b 9B | 88.2% | huggingface.co | Official leaderboard | ||
| 2 | perplexity-ai/pplx-embed-v2-late-0.6b 0.6B | 86.7% | huggingface.co | Official leaderboard | ||
| 3 | perplexity-ai/pplx-embed-v1-4b 4B | 85% | huggingface.co | Official leaderboard | ||
| 4 | nvidia/Nemotron-3-Embed-8B-BF16 8B | 83.1% | huggingface.co | Official leaderboard | ||
| 5 | perplexity-ai/pplx-embed-v1-0.6b 0.6B | 81.8% | huggingface.co | Official leaderboard | ||
| 6 | google/embeddinggemma-300m 0.3B | 80.1% | huggingface.co | Official leaderboard | ||
| 7 | | 79.3% | Jun 2025 | huggingface.co | Official leaderboard | |
| 8 | topk-io/topk-embed-v1-small (128 dim output) 2.2B | 79.2% | huggingface.co | Official leaderboard | ||
| 9 | perplexity-ai/pplx-embed-v1-late-0.6b 0.6B | 78.6% | huggingface.co | Official leaderboard | ||
| 10 | topk-io/topk-embed-v1-xsmall (128 dim output) 0.8B | 77.4% | huggingface.co | Official leaderboard | ||
| 11 | | 76.5% | Jun 2025 | huggingface.co | Official leaderboard | |
| 12 | nvidia/Nemotron-3-Embed-1B-BF16 1.1B | 75.6% | huggingface.co | Official leaderboard | ||
| 13 | jinaai/jina-embeddings-v5-text-small 0.6B | 74.8% | huggingface.co | Official leaderboard | ||
| 14 | lightonai/mLateOn 0.3B | 74.2% | huggingface.co | Official leaderboard | ||
| 15 | lightonai/mDenseOn 0.3B | 73.7% | huggingface.co | Official leaderboard | ||
| 16 | chungimungi/GLInt 0.2B | 72.6% | huggingface.co | Official leaderboard | ||
| 17 | | 71.2% | Jun 2025 | huggingface.co | Official leaderboard | |
| 18 | voyageai/voyage-4-nano 0.3B | 69.5% | huggingface.co | Official leaderboard | ||
| 19 | BM25-tantivy | 60.8% | github.com | Official leaderboard |
| # | Model | Score | Setting | Date | Link | Source |
|---|---|---|---|---|---|---|
| 1 | | 79.3% | Jun 2025 | huggingface.co | Official leaderboard | |
| 2 | | 76.5% | Jun 2025 | huggingface.co | Official leaderboard | |
| 3 | | 71.2% | Jun 2025 | huggingface.co | Official leaderboard |
Rankings by part
We read 94 parts from the benchmark’s own leaderboard in 2 groups: full corpus and subcorpus. Each ranks models on one part of the test. Qwen3-Embedding-8B leads every one of them. Arrows show how far a model moved from its overall rank among the models we track.
R@1000 Citation
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 79.3%
- 2
Qwen3-Embedding-4B 76.5%
- 3
Qwen3-Embedding-0.6B 71.2%
R@1000 Web Ranking
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 58.3%
- 2
Qwen3-Embedding-4B 55.6%
- 3
Qwen3-Embedding-0.6B 51.8%
R@1000 Comb.
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 65.3%
- 2
Qwen3-Embedding-4B 62.8%
- 3
Qwen3-Embedding-0.6B 58.7%
R@100 Citation
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 47.3%
- 2
Qwen3-Embedding-4B 44%
- 3
Qwen3-Embedding-0.6B 39.4%
R@100 Web Ranking
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 24.1%
- 2
Qwen3-Embedding-4B 22.3%
- 3
Qwen3-Embedding-0.6B 20.6%
R@100 Comb.
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 26.9%
- 2
Qwen3-Embedding-4B 25.2%
- 3
Qwen3-Embedding-0.6B 22.8%
nDCG@10 Citation
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 10.8%
- 2
Qwen3-Embedding-4B 9.91%
- 3
Qwen3-Embedding-0.6B 8.53%
nDCG@10 Web Ranking
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 16.3%
- 2
Qwen3-Embedding-4B 15%
- 3
Qwen3-Embedding-0.6B 14.1%
nDCG@10 Comb.
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 41.2%
- 2
Qwen3-Embedding-4B 38.8%
- 3
Qwen3-Embedding-0.6B 36.2%
R@1000 Comb. · English
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 66.2%
- 2
Qwen3-Embedding-4B 63%
- 3
Qwen3-Embedding-0.6B 59.6%
R@1000 Comb. · Spanish
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 61.4%
- 2
Qwen3-Embedding-4B 60%
- 3
Qwen3-Embedding-0.6B 55.2%
R@1000 Comb. · Russian
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 64.1%
- 2
Qwen3-Embedding-4B 63.2%
- 3
Qwen3-Embedding-0.6B 58%
R@1000 Comb. · German
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 61.4%
- 2
Qwen3-Embedding-4B 60.5%
- 3
Qwen3-Embedding-0.6B 54.1%
R@1000 Comb. · French
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 65.2%
- 2
Qwen3-Embedding-4B 64.3%
- 3
Qwen3-Embedding-0.6B 57.7%
R@1000 Comb. · Portuguese
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 67.1%
- 2
Qwen3-Embedding-4B 65.7%
- 3
Qwen3-Embedding-0.6B 57.6%
R@1000 Comb. · Italian
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 65.6%
- 2
Qwen3-Embedding-4B 63%
- 3
Qwen3-Embedding-0.6B 57.2%
R@1000 Comb. · Korean
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 62.5%
- 2
Qwen3-Embedding-4B 60.9%
- 3
Qwen3-Embedding-0.6B 59.1%
R@1000 Comb. · Japanese
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 66.9%
- 2
Qwen3-Embedding-4B 65.6%
- 3
Qwen3-Embedding-0.6B 60.3%
R@1000 Comb. · Chinese
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 61.5%
- 2
Qwen3-Embedding-4B 60.6%
- 3
Qwen3-Embedding-0.6B 56.6%
R@1000 Comb. · Consumer Goods
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 66.3%
- 2
Qwen3-Embedding-4B 63.1%
- 3
Qwen3-Embedding-0.6B 58.2%
R@1000 Comb. · Programming
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 65.9%
- 2
Qwen3-Embedding-4B 61.4%
- 3
Qwen3-Embedding-0.6B 58.5%
R@1000 Comb. · Law
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 65.9%
- 2
Qwen3-Embedding-4B 64.4%
- 3
Qwen3-Embedding-0.6B 58.4%
R@1000 Comb. · Health
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 67.1%
- 2
Qwen3-Embedding-4B 64.8%
- 3
Qwen3-Embedding-0.6B 60.8%
R@1000 Comb. · Business
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 61.9%
- 2
Qwen3-Embedding-4B 58.9%
- 3
Qwen3-Embedding-0.6B 55.7%
R@1000 Comb. · Science
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 67.7%
- 2
Qwen3-Embedding-4B 65.1%
- 3
Qwen3-Embedding-0.6B 62.5%
R@1000 Comb. · Education
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 61.3%
- 2
Qwen3-Embedding-4B 59.1%
- 3
Qwen3-Embedding-0.6B 54.5%
R@1000 Comb. · Technology
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 65.2%
- 2
Qwen3-Embedding-4B 62.1%
- 3
Qwen3-Embedding-0.6B 59.1%
R@1000 Comb. · Finance
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 64%
- 2
Qwen3-Embedding-4B 62%
- 3
Qwen3-Embedding-0.6B 58.1%
R@1000 Comb. · Real Estate
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 66.6%
- 2
Qwen3-Embedding-4B 64.7%
- 3
Qwen3-Embedding-0.6B 59.6%
R@1000 Comb. · Writing
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 65.4%
- 2
Qwen3-Embedding-4B 63.1%
- 3
Qwen3-Embedding-0.6B 59.9%
R@1000 Comb. · Travel
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 68.4%
- 2
Qwen3-Embedding-4B 66.4%
- 3
Qwen3-Embedding-0.6B 61.6%
R@1000 Comb. · Entertainment
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 65.8%
- 2
Qwen3-Embedding-4B 64%
- 3
Qwen3-Embedding-0.6B 58.2%
R@1000 Comb. · Careers
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 64.8%
- 2
Qwen3-Embedding-4B 62.6%
- 3
Qwen3-Embedding-0.6B 58.3%
R@1000 Comb. · Politics
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 65.1%
- 2
Qwen3-Embedding-4B 63.6%
- 3
Qwen3-Embedding-0.6B 59.8%
R@1000 Comb. · Local
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 61.1%
- 2
Qwen3-Embedding-4B 60%
- 3
Qwen3-Embedding-0.6B 53.6%
R@1000 Comb. · Food
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 71.1%
- 2
Qwen3-Embedding-4B 68.8%
- 3
Qwen3-Embedding-0.6B 64.8%
R@1000 Comb. · Social Media
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 66.5%
- 2
Qwen3-Embedding-4B 62.5%
- 3
Qwen3-Embedding-0.6B 59.1%
R@1000 Comb. · Environment
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 67.9%
- 2
Qwen3-Embedding-4B 66.3%
- 3
Qwen3-Embedding-0.6B 62.8%
R@1000 Comb. · Art
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 66.3%
- 2
Qwen3-Embedding-4B 64.2%
- 3
Qwen3-Embedding-0.6B 61%
R@1000 Comb. · News
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 53.4%
- 2
Qwen3-Embedding-4B 51.5%
- 3
Qwen3-Embedding-0.6B 47.6%
R@1000 Comb. · History
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 71.5%
- 2
Qwen3-Embedding-4B 69.4%
- 3
Qwen3-Embedding-0.6B 66.8%
R@1000 Comb. · Other
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 65.8%
- 2
Qwen3-Embedding-4B 63%
- 3
Qwen3-Embedding-0.6B 59.5%
R@1000 Comb. · People
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 65%
- 2
Qwen3-Embedding-4B 63.5%
- 3
Qwen3-Embedding-0.6B 59.8%
R@1000 Comb. · Role Play
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 63.1%
- 2
Qwen3-Embedding-4B 60.2%
- 3
Qwen3-Embedding-0.6B 54.6%
R@1000 Comb. · Translation
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 58.9%
- 2
Qwen3-Embedding-4B 56.1%
- 3
Qwen3-Embedding-0.6B 54.3%
R@1000 Comb. · Support
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 64.9%
- 2
Qwen3-Embedding-4B 62.3%
- 3
Qwen3-Embedding-0.6B 58.1%
R@1000 Comb. · Primary
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 67.1%
- 2
Qwen3-Embedding-4B 64.9%
- 3
Qwen3-Embedding-0.6B 61.3%
R@1000 Citation
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 82.7%
- 2
Qwen3-Embedding-4B 80.5%
- 3
Qwen3-Embedding-0.6B 75.6%
R@1000 Web Ranking
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 63.6%
- 2
Qwen3-Embedding-4B 61.4%
- 3
Qwen3-Embedding-0.6B 57.4%
R@1000 Comb.
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 70.4%
- 2
Qwen3-Embedding-4B 68.4%
- 3
Qwen3-Embedding-0.6B 64.4%
R@100 Citation
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 48.1%
- 2
Qwen3-Embedding-4B 44.8%
- 3
Qwen3-Embedding-0.6B 40.1%
R@100 Web Ranking
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 24.7%
- 2
Qwen3-Embedding-4B 23%
- 3
Qwen3-Embedding-0.6B 21.2%
R@100 Comb.
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 27.6%
- 2
Qwen3-Embedding-4B 26%
- 3
Qwen3-Embedding-0.6B 23.4%
nDCG@10 Citation
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 10.8%
- 2
Qwen3-Embedding-4B 9.95%
- 3
Qwen3-Embedding-0.6B 8.57%
nDCG@10 Web Ranking
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 16.4%
- 2
Qwen3-Embedding-4B 15.1%
- 3
Qwen3-Embedding-0.6B 14.2%
nDCG@10 Comb.
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 41.4%
- 2
Qwen3-Embedding-4B 39.1%
- 3
Qwen3-Embedding-0.6B 36.4%
R@1000 Comb. · English
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 70.8%
- 2
Qwen3-Embedding-4B 68.3%
- 3
Qwen3-Embedding-0.6B 65%
R@1000 Comb. · Spanish
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 66.5%
- 2
Qwen3-Embedding-4B 65.1%
- 3
Qwen3-Embedding-0.6B 60.4%
R@1000 Comb. · Russian
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 70.7%
- 2
Qwen3-Embedding-4B 70%
- 3
Qwen3-Embedding-0.6B 65%
R@1000 Comb. · German
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 67.7%
- 2
Qwen3-Embedding-4B 67%
- 3
Qwen3-Embedding-0.6B 60.8%
R@1000 Comb. · French
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 71.6%
- 2
Qwen3-Embedding-4B 70.8%
- 3
Qwen3-Embedding-0.6B 64.6%
R@1000 Comb. · Portuguese
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 72.4%
- 2
Qwen3-Embedding-4B 71.2%
- 3
Qwen3-Embedding-0.6B 63.6%
R@1000 Comb. · Italian
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 71%
- 2
Qwen3-Embedding-4B 68.7%
- 3
Qwen3-Embedding-0.6B 63.1%
R@1000 Comb. · Korean
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 69.9%
- 2
Qwen3-Embedding-4B 68.8%
- 3
Qwen3-Embedding-0.6B 66.7%
R@1000 Comb. · Japanese
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 74.3%
- 2
Qwen3-Embedding-4B 73.1%
- 3
Qwen3-Embedding-0.6B 67.3%
R@1000 Comb. · Chinese
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 68.6%
- 2
Qwen3-Embedding-4B 67.7%
- 3
Qwen3-Embedding-0.6B 63.2%
R@1000 Comb. · Consumer Goods
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 71%
- 2
Qwen3-Embedding-4B 68.5%
- 3
Qwen3-Embedding-0.6B 63.8%
R@1000 Comb. · Programming
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 70.8%
- 2
Qwen3-Embedding-4B 67.1%
- 3
Qwen3-Embedding-0.6B 64.5%
R@1000 Comb. · Law
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 70.3%
- 2
Qwen3-Embedding-4B 68.9%
- 3
Qwen3-Embedding-0.6B 63.3%
R@1000 Comb. · Health
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 72.3%
- 2
Qwen3-Embedding-4B 70.4%
- 3
Qwen3-Embedding-0.6B 66.4%
R@1000 Comb. · Business
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 67.8%
- 2
Qwen3-Embedding-4B 65.5%
- 3
Qwen3-Embedding-0.6B 62.1%
R@1000 Comb. · Science
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 73.2%
- 2
Qwen3-Embedding-4B 71.2%
- 3
Qwen3-Embedding-0.6B 68.4%
R@1000 Comb. · Education
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 66.5%
- 2
Qwen3-Embedding-4B 64.7%
- 3
Qwen3-Embedding-0.6B 60.1%
R@1000 Comb. · Technology
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 70.7%
- 2
Qwen3-Embedding-4B 68.3%
- 3
Qwen3-Embedding-0.6B 65.4%
R@1000 Comb. · Finance
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 69%
- 2
Qwen3-Embedding-4B 67.4%
- 3
Qwen3-Embedding-0.6B 63.6%
R@1000 Comb. · Real Estate
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 71.5%
- 2
Qwen3-Embedding-4B 69.7%
- 3
Qwen3-Embedding-0.6B 65%
R@1000 Comb. · Writing
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 70.9%
- 2
Qwen3-Embedding-4B 69.2%
- 3
Qwen3-Embedding-0.6B 65.5%
R@1000 Comb. · Travel
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 73.1%
- 2
Qwen3-Embedding-4B 71.4%
- 3
Qwen3-Embedding-0.6B 66.9%
R@1000 Comb. · Entertainment
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 70.3%
- 2
Qwen3-Embedding-4B 69.2%
- 3
Qwen3-Embedding-0.6B 63.6%
R@1000 Comb. · Careers
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 70.1%
- 2
Qwen3-Embedding-4B 68.4%
- 3
Qwen3-Embedding-0.6B 64.3%
R@1000 Comb. · Politics
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 69.8%
- 2
Qwen3-Embedding-4B 68.8%
- 3
Qwen3-Embedding-0.6B 65%
R@1000 Comb. · Local
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 67%
- 2
Qwen3-Embedding-4B 66.2%
- 3
Qwen3-Embedding-0.6B 59.8%
R@1000 Comb. · Food
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 76.4%
- 2
Qwen3-Embedding-4B 75%
- 3
Qwen3-Embedding-0.6B 70.8%
R@1000 Comb. · Social Media
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 71.3%
- 2
Qwen3-Embedding-4B 67.8%
- 3
Qwen3-Embedding-0.6B 64.7%
R@1000 Comb. · Environment
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 72.8%
- 2
Qwen3-Embedding-4B 71.8%
- 3
Qwen3-Embedding-0.6B 68.2%
R@1000 Comb. · Art
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 71.8%
- 2
Qwen3-Embedding-4B 70.4%
- 3
Qwen3-Embedding-0.6B 66.7%
R@1000 Comb. · News
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 58.6%
- 2
Qwen3-Embedding-4B 57.1%
- 3
Qwen3-Embedding-0.6B 53.1%
R@1000 Comb. · History
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 77%
- 2
Qwen3-Embedding-4B 75.6%
- 3
Qwen3-Embedding-0.6B 72.7%
R@1000 Comb. · Other
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 71.2%
- 2
Qwen3-Embedding-4B 69.3%
- 3
Qwen3-Embedding-0.6B 65.4%
R@1000 Comb. · People
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 70.3%
- 2
Qwen3-Embedding-4B 69%
- 3
Qwen3-Embedding-0.6B 64.9%
R@1000 Comb. · Role Play
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 69%
- 2
Qwen3-Embedding-4B 67%
- 3
Qwen3-Embedding-0.6B 60.8%
R@1000 Comb. · Translation
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 65.7%
- 2
Qwen3-Embedding-4B 63.8%
- 3
Qwen3-Embedding-0.6B 61%
R@1000 Comb. · Support
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 70%
- 2
Qwen3-Embedding-4B 67.8%
- 3
Qwen3-Embedding-0.6B 63.7%
R@1000 Comb. · Primary
3 models we track have a score here.
- 1
Qwen3-Embedding-8B 72.7%
- 2
Qwen3-Embedding-4B 71%
- 3
Qwen3-Embedding-0.6B 67.4%
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official Q2D-Web leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official Q2D-Web leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
More in Knowledge
- MMLU-Pro In use Harder 10-option version of MMLU with reasoning-focused questions across 14 subjects.
Intern-S2-397B 89.8%
- MMLU In use 57-subject multiple-choice exam questions from elementary to professional level. Scored by accuracy.
Seed1.8 92.3%
- SuperGPQA In use 26,529 graduate-level multiple-choice questions across 285 disciplines. Accuracy.
Qwen 3.6 Max (Preview) 73.9%
- WideSearch In use 200 bilingual broad info-seeking tasks: agent collects many atomic facts into a table. Scored by success rate and row/item F1.
Tencent Hy4 preview 83.9%
- BBEH micro average In use Models solve harder replacements for each of the 23 BIG-Bench Hard tasks, 4,520 examples in total. The score is the share of all examples answered correctly.
Gemma 4 31B IT 74.4%
- HLE-Diamond In use Models answer 1,000 cleaned expert questions from Humanity's Last Exam, 500 on reasoning and 500 on knowledge, closed book. Graded on exact answers. Claude Fable 5.2 77.9%
Data audit
9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 17 months ago, in June 2025, for Qwen3-Embedding-0.6B. Every benchmark
Catalog record 2 pass
- Pass Task access The tasks are private.
- Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
- Pass Independent runs Scores from the official leaderboard (19 entries).
- Pass Models scored 3 models we track, plus 16 other entries on its leaderboard.
- Warning Newest score The newest came 17 months ago, in June 2025, for Qwen3-Embedding-0.6B.
- Note Frontier models None of its models are on the capabilities index.
- Pass Headroom The best is 79.3%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on huggingface.co: 3 rows for models we track and 16 other entries at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.