BreadBowl-Embed v1.3
One model. As good as two. It encodes each passage once into 16 slots, then finds and reranks candidates from those same stored vectors. No cross-encoder. No second pass over the text.
Same candidates for both (v1.3's, 72 per query on average), one RTX 5090. The cross-encoder is Qwen3-Reranker-0.6B; its time comes from measured batch throughput. ▼ significantly below v1.3.
- As good as two models. Level with Qwen3-Embedding-0.6B + Qwen3-Reranker-0.6B on MS MARCO and DL19, and 5.5 nDCG@10 ahead on DL20.
- 50× faster rerank. 2.9 ms against 150 ms on the same candidates.
- Ahead of its first teacher on DL20. 3.7 points above Qwen3-Reranker-0.6B on the same candidates; level on MS MARCO and DL19. No candidate text is read at query time.
- 6 wins, 9 level, 0 losses against released models of similar size.
- BEIR: 55.1 mean over 15 tasks, 3.6 above v1.2-preview.
📄 Paper: BreadBowl-Embed: Encode Once, Retrieve and Rerank
📝 Release post: One model. As good as two.
How it works
- Encode once. A 0.8B backbone reads each passage once, at index time, into 16 slots. At search time only the query runs through the model.
- Find. Each slot holds a routing vector and a value vector. The query's routing vectors search the stored routing vectors.
- Rerank. The same routing match decides which stored value vectors the query reads. That is a few small matrix products, not a model pass.
- Score = 0.75 × routing + 0.25 × value. The toolkit splits every score into slot-pair parts you can inspect.
- Fixed size. 16 + 16 vectors per passage, however long it is. At int4: 4,160 bytes, for at most 0.12 points.
Quick start
Python 3.12 or newer. Install the library from GitHub:
pip install "git+https://github.com/BreadBowlAI/breadbowl-embed.git"
from breadbowl_embed import BreadBowl
model = BreadBowl.from_pretrained("dotproductx/breadbowl-embed-v1.3")
ids = ["octopus", "ocean", "plants"]
passages = ["Octopuses have three hearts.",
"The Pacific is Earth's largest ocean.",
"Plants use sunlight during photosynthesis."]
index = model.index(model.encode_documents(passages), ids=ids) # encode once, keep
query = model.encode_queries(["How many hearts does an octopus have?"])
for hit in index.search(query, top_k=2)[0]: # retrieve, then rerank
print(hit.id, round(hit.score, 3))
More examples (rerank a list, save an index, explain a score) are in the library README.
Results
| nDCG@10, full MS MARCO corpus | Models | MS MARCO dev | DL19 | DL20 |
|---|---|---|---|---|
| v1.3 | 1 | 41.7 | 74.6 | 73.1 |
| Qwen3-Embedding-0.6B + Qwen3-Reranker-0.6B | 2 | 41.9 | 72.0 | 67.6 ▼ |
| Jina v5 text-small + Qwen3-Reranker-0.6B | 2 | 42.0 | 72.6 | 66.8 ▼ |
| Qwen3-Embedding-0.6B alone | 1 | 38.0 ▼ | 66.8 ▼ | 67.6 ▼ |
| v1.3 + Qwen3-Reranker-4B | 2 | 44.6 ▲ | 75.3 | 75.3 |
▼ / ▲ significantly below / above v1.3 (95% paired bootstrap); unmarked rows are level with it. Every setting is in the paper.
Limits
- English only. MS MARCO is in the training data. DL19 and DL20 have 43 and 54 queries. Trained once, not repeated over seeds.
- Not top of every chart. Qwen3-Reranker-4B, five times larger, is ahead on MS MARCO. On BEIR, the Jina v5 paper reports higher means for Qwen3-Embedding-0.6B and Jina v5 text-small, under other settings. On argument retrieval the value score falls short of the cross-encoders.
- CPU is slow. About 70 ms per query and 2.7 passages per second on a 16-core desktop CPU (measured with the library). CUDA is not yet validated in the library.
- Precision. Results were measured in bfloat16; load with
dtype="bfloat16"to match. - Scores rank candidates; they are not probabilities. This is a research release.
- Licence. Weights: Apache-2.0. Base model: Qwen/Qwen3.5-0.8B-Base. Training data keeps its own terms (MS MARCO's describe non-commercial research use); check them before you use the model.
Links
- Paper: paper/breadbowl-embed-v1.3.pdf
- Release post: One model. As good as two.
- Code: github.com/BreadBowlAI/breadbowl-embed
- Full details (BEIR per task, storage, speed, training data, citation): model-details-v1.3.md
- Previous version: v1.2-preview
- Downloads last month
- 111
Model tree for dotproductx/breadbowl-embed-v1.3
Base model
Qwen/Qwen3.5-0.8B-BaseDatasets used to train dotproductx/breadbowl-embed-v1.3
lightonai/embeddings-pre-training-curated
allenai/scirepeval
Evaluation results
- nDCG@10, QKV (0.75 routing + 0.25 value), slot-hit width 128 on MS MARCO passage dev (BEIR, 6,980 queries, full 8,841,823-passage corpus)self-reported41.710
- nDCG@10, QK (routing only), slot-hit width 128 on MS MARCO passage dev (BEIR, 6,980 queries, full 8,841,823-passage corpus)self-reported39.300
- Recall@100, slot-hit width 128, relevance >= 1 on MS MARCO passage dev (BEIR, 6,980 queries, full 8,841,823-passage corpus)self-reported84.460
- nDCG@10, QKV (0.75 routing + 0.25 value), slot-hit width 128 on TREC DL 2019 passage (43 judged queries, same 8,841,823-passage corpus)test set self-reported74.630
- nDCG@10, QK (routing only), slot-hit width 128 on TREC DL 2019 passage (43 judged queries, same 8,841,823-passage corpus)test set self-reported71.560
- Recall@100, slot-hit width 128, relevance >= 2 on TREC DL 2019 passage (43 judged queries, same 8,841,823-passage corpus)test set self-reported57.010
- nDCG@10, QKV (0.75 routing + 0.25 value), slot-hit width 128 on TREC DL 2020 passage (54 judged queries, same 8,841,823-passage corpus)test set self-reported73.140
- nDCG@10, QK (routing only), slot-hit width 128 on TREC DL 2020 passage (54 judged queries, same 8,841,823-passage corpus)test set self-reported68.730

