The story so far
Weights, every dated article, and repo commits on one rail.
11 repo commits since the weights landed
Teal is the weights (2023-09-12), grey ticks are repo commits.
bge-small-zh-v1.5
24.0M parameters, text in, embeddings out, 512 tokens of context. 1 provider serves it.
Verified 2026-10-25 Rank #30 in Embeddings · score 94.3 method v1 Sources 3 Downloads 30d 3.5M · #41 of 619 tracked
Do people use it?3.5Mdownloads, 30 days#41 of 619 we track · top 7%+6 likes, downloads −8% this week162 likesHugging Face2026-10-24
Where to run it
Live hosted endpoints, cheapest first, USD per million tokens. Ties break on output price, then provider name.
| # | Cheapest live row | In /1M | Out /1M | Context | Quant | Seen |
|---|---|---|---|---|---|---|
| 1 | HF Inference | — | — | — | — | Oct 25 |
Where it's used
Hugging Face does not say who downloads a model or with what. These are the proxies: API providers listing it, quantized builds, fine-tunes/adapters/merges, and Spaces that load it.
Served by 1 API provider; 10 quantized builds for local use.
- As a service1HF Inference
- On own hardware10quantized builds
- Built on1111 fine-tunes · 0 adapters · 0 merges
- In apps100+Hugging Face Spaces
Top Spaces running it: embedl/hfviewer ♥ 181 · lightmate/llm-chatbot ♥ 9 · emanforhelp/model-pulse ♥ 5 · silk-road/ChatHaruhi-Needy ♥ 3
Will it fit?
Weights only, an estimate: leave room for the KV cache. Method
Fits a 16 GB card at BF16
About 0.1 GB at BF16: weights only, so leave room for the KV cache.
Estimate params × bytes per param × 1.2 · how we estimate
What people built on it
21 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.
21 derivatives on the Hub, most of them fine-tunes
From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.
Reference
Everything else, folded: the full tables, every source.
What the card says
1 cited sentence
Summary generated, cited
The model maps text to low-dimensional dense vectors for tasks such as retrieval, classification, or semantic search.[1]
Generated 2026-10-07 15:02 UTC by parasail-gemma-4-26b-a4b-it; each numeral points at the fact it rests on. Not a verdict.
Use-case evidence
2 use cases
What it is good at 2 use cases
Labels are evidence, not verdicts: measured a third-party leaderboard row for this model; author-reported a number the model author published in the card or eval files; structural a deterministic signal from the name, config or chat template, or a decision-model estimate. Each chip links to its source. All use cases.
Facts
21 fields
Listed in Embeddings
Facts
- bge-small-zh-v1.5 has 24.0M parameters. (safetensors metadata, observed 2026-10-24)
- Context length is 512 tokens (512). (config.json max_position_embeddings, observed 2026-10-24)
- License: MIT (friction: none). Permissive. (license file, observed 2026-10-24)
- Created 2023-09-12, last modified 2023-10-12. (HF API, observed 2026-10-24)
| Pipeline tag | feature-extraction |
|---|---|
| Library | transformers |
| Architectures | BertModel |
| Model type | bert |
| Dtype | float32 |
| Params by dtype | F32: 23,953,920 · I64: 512 |
| Size band | S |
| MoE | no |
| Context length | 512 (config.json max_position_embeddings) |
| Languages | zh |
| Papers | arXiv:2310.07554, arXiv:2309.07597 |
| License | MIT (mit) |
| Hub snapshot | 2026-10-25 |
Servability
vLLM · SGLang
Servability
vLLM: supported · BertModel · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · BertModel · 773cb938fe9484a8f7e7e46fb1b1fe96412c59ed
vLLM registry (7050ef881ff0835f6890a47db1859a24c65481d6) · SGLang registry (773cb938fe9484a8f7e7e46fb1b1fe96412c59ed) · Matched by architecture class name against each project's registry on main; 'supported' means the architecture is registered, not that this checkpoint was tested.
Hardware fit
3 estimates
Hardware fit ESTIMATE
Weights-only memory for bge-small-zh-v1.5: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 0.1 GB · Edge (<4 GB) |
|---|---|
| FP8 (1 byte/param) | 0.0 GB · Edge (<4 GB) |
| INT4 (0.5 bytes/param) | 0.0 GB · Edge (<4 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.
All variants
21 variants
Lineage
quantized: 10finetune: 11
Top variants (21 shown, ranked by downloads)
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| Xenova/bge-small-zh-v1.5 | quantized | — | — | — | — | Xenova | 6 | 291k |
| onnx-community/bge-small-zh-v1.5-ONNX | quantized | — | — | — | — | onnx-community | 1 | 7.5k |
| yenstdi/ntu-bge-small-zh-simcse-job-talent-matching | finetune | — | 24.0M | S | — | yenstdi | 0 | 1.1k |
| johnyy212/moe-girl-v2 | finetune | — | 24.0M | S | — | johnyy212 | 0 | 522 |
| baicuya/fine_tuned_bge_small_zh_v1.5 | finetune | — | 24.0M | S | — | baicuya | 0 | 471 |
| ntucool/mlogging | finetune | — | 24.0M | S | — | ntucool | 1 | 387 |
| johnyy212/moe-girl-v3 | finetune | — | 24.0M | S | — | johnyy212 | 0 | 364 |
| johnyy212/moe-girl-test | finetune | — | 24.0M | S | — | johnyy212 | 0 | 312 |
| dumyy/my_awesome_model | finetune | — | 24.0M | S | mit | dumyy | 0 | 264 |
| JoramMillenaar/bge-small-zh-v1.5-vocab-quantized | quantized | -quant | — | — | mit | JoramMillenaar | 0 | 224 |
| infinilabs/pizza-embed-zh-small-v1 | quantized | — | — | — | mit | infinilabs | 1 | 155 |
| zhufucdev/BAAI-bge-small-zh-v1.5 | quantized | — | — | — | apache-2.0 | zhufucdev | 0 | 196 |
| JasonYANG170/bge-small-zh-v1.5-ONNX | quantized | — | — | — | mit | JasonYANG170 | 0 | 54 |
| dumyy/sft-bge-bert24m-to-sentiment | finetune | — | 24.0M | S | mit | dumyy | 0 | 49 |
| wybxc/bert-small-mygo | finetune | — | 24.0M | S | — | wybxc | 0 | 49 |
| vteaw/bge-small-zh-v1.5 | quantized | — | — | — | — | vteaw | 0 | 47 |
| zjtakato/bge-small-zh-v1.5 | quantized | — | — | — | — | zjtakato | 0 | 46 |
| NCHUPPBL/nchu-bge-small-v2 | finetune | — | 24.0M | S | — | NCHUPPBL | 0 | 27 |
| jatshi/StreamSense-Serve-v4-Router | finetune | — | — | — | — | jatshi | 0 | 0 |
| GoForceX/roleplay-quality-onnx | quantized | — | — | — | mit | GoForceX | 0 | 0 |
| GoForceX/roleplay-jev-v2-int8 | quantized | int8 | — | — | apache-2.0 | GoForceX | 0 | 0 |
Repository changelog
11 commits
Repository changelog 11 commits
| Date | Commit | SHA |
|---|---|---|
| 2023-10-12 | update | 7999e1d |
| 2023-09-27 | Merge branch 'main' of https://huggingface.co/BAAI/bge-small-zh-v1.5 into main | a7ec183 |
| 2023-09-27 | update | e534609 |
| 2023-09-18 | Adding `safetensors` variant of this model (#1) | 2647854 |
| 2023-09-15 | update | 13942ee |
| 2023-09-12 | update | 0cfcd48 |
| 2023-09-12 | update | 082b8e2 |
| 2023-09-12 | update | 4e17e24 |
| 2023-09-12 | update | c9c382e |
| 2023-09-12 | upload | 4bf3c54 |
| 2023-09-12 | initial commit | 6e72d86 |
All providers
1 provider · 1 endpoint
Where to run 1 providers · 1 endpoints
Hosted endpoints observed for bge-small-zh-v1.5, sorted by input price (USD per million tokens). Each row names its observer and date; prices change daily, so check the source before quoting.
| Provider | Model ref | Context | Quant | $/M in | $/M out | tok/s | TTFT | Tools | Status | Observed |
|---|---|---|---|---|---|---|---|---|---|---|
| HF Inference hf | BAAI/bge-small-zh-v1.5 | — | — | — | — | — | — | — | live | 2026-10-25 |
Articles and coverage
7 pieces
Ranked by source signals (HN, Lobsters, authority domains, official posts) and recency. Summaries are generated from a stored copy of the article text; every sentence cites a span of that text, and nothing is shown that failed the citation check.
Analysis6
github.comauthority819 words
A local service wraps the BAAI/bge-small-zh-v1.5 model to provide an OpenAI-compatible API for Chinese text embedding. The model outputs 512-dimensional vectors and requires approximately 80MB of memory. It is released under the Apache 2.0 license and runs on port 7078.written, reviewed, cited
- The service supports batch embedding requests by accepting an array of strings in the input parameter.
- The OpenClaw configuration uses a placeholder for the API key because the local service does not require one.
hfviewer.com3,316 words
hfviewer renders an interactive architecture graph for the Hugging Face model BAAI/bge-small-zh-v1.5, displaying nodes for modules like embeddings and attention blocks with edges following the actual forward dataflow. The model uses 4 transformer layers, a hidden size of 512, and 8 attention heads, with feed-forward layers using an intermediate size of 2,048. It features a vocabulary of 21,128 tokens, a context window of up to 512 tokens, and checkpoint weights stored in F32.written, reviewed, cited
- The graph is trace-backed where executable, otherwise derived from reviewed configuration and source code.
- hfviewer allows users to paste a Hugging Face model URL to open an interactive architecture graph in the browser.
- Family pages like Gemma 4 enable comparing multiple related models with synchronized interaction instead of isolated screenshots.
ollama.com241 words
The model features a block_count of 4 and an attention head_count of 8. It supports a context_length of 512 tokens with an embedding_length of 512.generated, cited
adg.csdn.net718 words
BAAI released the bge-small-zh-v1.5 model family to solve challenges in Chinese semantic retrieval where large models are resource-heavy and small models lack precision. The family includes three versions tested on 31 datasets in the C-MTEB benchmark, with the small variant achieving 86.2% Top-10 accuracy. Engineers can reduce model size by 75% using INT8 quantization to improve inference speed for resource-constrained scenarios.written, reviewed, cited
- Using INT8 quantization reduces the model volume by 75% and increases inference speed by 2-3 times.
- Engineers should update the model when accuracy drops by more than 10% or when new domain data is available.
dataloop.ai
dataloop.aiabout the bge small family
Community1
Papers
2 papers
Papers 2
arXiv links found in the model card and its tags. "Peer-reviewed" is shown only with evidence: a published venue in OpenAlex (measured) or the authors' own journal reference on arXiv (reported).
- A Multi-Task Embedder For Retrieval Augmented LLMs cited in card
- C-Pack: Packed Resources For General Chinese Embeddings cited in card
Similar models
3 models
Similar models
- bge-small-en-v1.5 · 33.4M same size band (S), same task (feature-extraction), permissive license
- jina-embeddings-v2-small-en · 32.7M same size band (S), same task (feature-extraction), permissive license
- granite-embedding-small-english-r2 · 47.7M same size band (S), same task (feature-extraction), permissive license
Sources
3 sources
Sources
Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:35 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:45 UTC · Summary: 2026-10-07 15:02 UTC · Last verified 2026-10-25
Follow bge-small-zh-v1.5
Watch this model
- /feeds/models/BAAI--bge-small-zh-v1.5.xml(RSS 2.0) · every change to bge-small-zh-v1.5
- /feeds/families/BAAI--bge-small.xml(RSS 2.0) · every change across the bge-small family
Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.