Runs as a custom pluginEmbeddingVerified 2026-11-20
bge-small-en-v1.5
BAAI/bge-small-en-v1.5
bge-small-en-v1.5 is a text embedding model (MIT). Runs on Mixpeek when a template imports it from Hugging Face.
Mixpeek does not run it today. It can run as a custom plugin: the weights load without remote code.
Used byBring your own model
No demo for bge-small-en-v1.5 yet
A video, image, slideshow or playground goes here once we record one on our own data.
What it is good and bad at
- Best at
- Small English text vectors (33M params, 384 dims) that a template imports from the Hub by repo and revision, so a namespace runs a model the user picked.
- Worst at
- Non-English text and passages over 512 tokens, which are truncated.
- Task
- Embedding
- Input
- Text
- Output
- Vector
- Use cases
- Text search
Templates that use it
- Bring your own modelDeploy this template
Name a Hugging Face model in the template and deploy it in one click. Apply imports the model, serves it, embeds your documents with it and gives you a search over them, with no upload step and no serving code.
Input and output
Input
Text
- Max tokens
- 512
- Truncation
- true
- Language
- English
Prompts
none
Output
How Mixpeek wires it
- Extractors
bge_small_hf@v1fixed (a manifest-declared extractor that extends text_extractor@v1 and names this model)
- Feature URIs
mixpeek://bge_small_hf@v1/embedding
- Vector indexes
- None
- Retriever stages
- feature_search
- Combines with
- a manifest extractor with source.huggingface (repo BAAI/bge-small-en-v1.5, a pinned revision) -> text vectors -> feature_search with a typed query
How it runs
- Status
- Runs as a custom plugin No Serve app or built-in extractor loads it. A manifest that declares it under extractors with source.huggingface imports it into the organization's model registry at apply (the bring-your-own-model template does).
- Serve apps
- None
- Runtime
- transformers AutoModel + AutoTokenizer (BertModel), CLS pooling
- Weights
- safetensors, no remote code
- Parameters
- 33.4 M
- Size on disk
- 133 MB
- Preprocessing
- Tokenize with truncation at 512 tokens.
- Postprocessing
- CLS token embedding, L2-normalized (384 dims).
- Serving
- both
- Hardware
- Runs on CPU (slower without a GPU)
Scale to zero
- Minimum replicas
- Not recorded
- Maximum replicas
- Not recorded
- Idle timeout
- Not recorded
- Memory
- Not recorded
- CPUs
- Not recorded
- GPUs
- Not recorded
- Cold start, node warm
- Not recorded
- Cold start, new node
- Not recorded
Measured cost
Unmeasured
No committed throughput or cost measurement.
Our evaluations
Unmeasured
We have not evaluated this model on our own data yet. Publisher results are in the sources below.
Questions
- What is bge-small-en-v1.5 best at?
- Small English text vectors (33M params, 384 dims) for search over short documents.
- What are the limits of bge-small-en-v1.5?
- Non-English text, and passages over 512 tokens, which are truncated.
- Can bge-small-en-v1.5 be used commercially?
- Yes. The weights are MIT licensed.
- How does bge-small-en-v1.5 run on Mixpeek?
- A manifest names it under extractors with source.huggingface and a pinned revision, and apply imports it into the organization's model registry. The bring-your-own-model template does this.
- What does bge-small-en-v1.5 output?
- A 384-dimensional vector per text, compared by cosine similarity.