SAVRN Model Hub · Comparisons
bge-reranker-large vs Qwen3-Embedding-0.6B
Bge-reranker-large has 560M parameters and Qwen3-Embedding-0.6B has 596M parameters; bge-reranker-large is released under MIT License and Qwen3-Embedding-0.6B under Apache License 2.0; at 16-bit, bge-reranker-large needs about 1.3 GB (1x MI300X from $1.85 an hour) and Qwen3-Embedding-0.6B about 1.4 GB (1x MI300X from $1.85 an hour).
| Field | bge-reranker-large BAAI/bge-reranker-large | Qwen3-Embedding-0.6B Qwen/Qwen3-Embedding-0.6B |
|---|---|---|
| Publisher | Beijing Academy of Artificial Intelligence | Qwen |
| Task | Feature extraction | Feature extraction |
| Modality | Text | Text |
| Parameters, as reported | 560M parameters | 596M parameters |
| Architecture | XLMRobertaForSequenceClassification | Qwen3ForCausalLM |
| Library | transformers | sentence-transformers |
| Context length | 514 tokens | 32,768 tokens |
| Repository size | 6.7 GB | 1.2 GB |
| Artifact formats | safetensors, onnx, pytorch | safetensors |
| License | mit | apache-2.0 |
| Access | Open weights, no gate | Open weights, no gate |
| Memory at 16-bit (weights and margin) | 1.3 GB | 1.4 GB |
| Cheapest GPUs at 16-bit, per hour | 1x MI300X, $1.85 | 1x MI300X, $1.85 |
| Memory at 4-bit (weights and margin) | 0.3 GB | 0.4 GB |
| Cheapest GPUs at 4-bit, per hour | 1x MI300X, $1.85 | 1x MI300X, $1.85 |
| Revision viewed | 55611d7bca2a | 97b0c614be4d |
| Downloads reported by the hub | 2.8M | 8.5M |
| Last observed | 2026-09-18 | 2026-09-18 |
An evaluation row appears only where at least two of these models report the same benchmark with the same stated configuration, metric, unit and setup. Different evaluators stay named in each cell. Values are shown as reported: no unit conversion, no ranking.
Other Reported Results
These results are listed for each model on its own, because the conditions needed to compare them are not stated or do not match. Two results that leave a condition blank are not assumed to share it.
bge-reranker-large
| Benchmark | Conditions | Result | Reported by | Revision | Date |
|---|---|---|---|---|---|
| MTEB CMedQAv1 | Configuration defaultTask RerankingMetric mapComparison conditions not established | 81.2721 | BAAI Publisher reported |
Evaluated revision not stated | — |
| MTEB CMedQAv1 | Configuration defaultTask RerankingMetric mrrComparison conditions not established | 84.1424 | BAAI Publisher reported |
Evaluated revision not stated | — |
| MTEB CMedQAv2 | Configuration defaultTask RerankingMetric mapComparison conditions not established | 84.1037 | BAAI Publisher reported |
Evaluated revision not stated | — |
| MTEB CMedQAv2 | Configuration defaultTask RerankingMetric mrrComparison conditions not established | 86.7938 | BAAI Publisher reported |
Evaluated revision not stated | — |
| MTEB MMarcoReranking | Configuration defaultTask RerankingMetric mapComparison conditions not established | 35.4601 | BAAI Publisher reported |
Evaluated revision not stated | — |
| MTEB MMarcoReranking | Configuration defaultTask RerankingMetric mrrComparison conditions not established | 34.6024 | BAAI Publisher reported |
Evaluated revision not stated | — |
| MTEB T2Reranking | Configuration defaultTask RerankingMetric mapComparison conditions not established | 67.2773 | BAAI Publisher reported |
Evaluated revision not stated | — |
| MTEB T2Reranking | Configuration defaultTask RerankingMetric mrrComparison conditions not established | 77.1315 | BAAI Publisher reported |
Evaluated revision not stated | — |
SAVRN's Notes on bge-reranker-large
Nobody gives a reranker its own accelerator, and at 1.3 GB in 16-bit bge-reranker-large shows why. This 560M-parameter XLM-RoBERTa sequence classifier from BAAI ranks retrieved passages for a retrieval-augmented pipeline, so we put it on the same card as that stack. On the cheapest listing the Index shows, a single MI300X with 192 GB at $1.85 an hour on-demand, it occupies under one percent of the memory.
MIT terms apply, so commercial use, modification and redistribution are allowed as long as the copyright and permission notices go with the files. The number to check is 514, the context length in tokens; whatever you hand it per call must fit that window, which sets chunk sizes upstream. The publisher's own page says it released newer rerankers with larger inputs and more languages on 3/18/2024; weigh that against this model's 2024-05-11 last update before standardizing.
SAVRN's Notes on Qwen3-Embedding-0.6B
Retrieval pipelines need an encoder that turns text into vectors, and this is the 0.6B of a series that also ships 4B and 8B. It needs 1.4 GB at 16-bit, 0.7 GB at 8-bit and 0.4 GB at 4-bit, so the cheapest setup, a 192 GB MI300X at $1.85 per hour, is a card it will share with the generator it feeds. The 32,768-token context matters more, since it sets how much of a document goes into one vector.
Apache 2.0 makes it a clean component to ship, with commercial use, modification and redistribution permitted, the license and NOTICE file kept and significant changes stated. It derives from Qwen3-0.6B-Base, and the architecture is Qwen3ForCausalLM served through sentence-transformers, so confirm your stack loads it that way. The Index lists no host price, so the comparison is your own accelerator time; arXiv:2506.05176 describes the method.
Questions
Which is larger, bge-reranker-large or Qwen3-Embedding-0.6B?
Qwen3-Embedding-0.6B (596M parameters) is larger than bge-reranker-large (560M parameters), by the parameter counts their publishers report.
Which is cheaper to run, bge-reranker-large or Qwen3-Embedding-0.6B?
At 4-bit, bge-reranker-large fits on 1x MI300X from $1.85 an hour and Qwen3-Embedding-0.6B on 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.
Can I use bge-reranker-large commercially?
Yes. bge-reranker-large is released under MIT License. The MIT License is a short permissive license. It permits commercial use, modification and redistribution, provided the copyright notice and permission notice are included.
Can I use Qwen3-Embedding-0.6B commercially?
Yes. Qwen3-Embedding-0.6B is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.