paraphrase-MiniLM-L6-v2 is a Sentence Transformers embedding model for sentences and short paragraphs. It maps each text into a 384-dimensional dense vector space, with semantic search and clustering among the maker’s stated tasks. The published architecture combines a transformer with mean pooling; its configured sequence length is 128 tokens.
Prepare the actual sentence or short paragraph, keeping the subject and important conditions intact. A practical collection can contain differently worded questions or brief descriptions of related topics. Embed each record and retain its original wording and identifier so a later comparison can return the text itself.
For example, several help questions may describe the same difficulty with different terms. A similarity stage can bring those candidates together, after which their wording can be inspected for the actual issue. The embedding is the numerical representation used in that comparison; it does not write a paraphrase or assign a final cluster label on its own.
The maker’s direct encoding example takes the attention mask into account when averaging token representations, excluding padding positions. Its architecture explicitly selects mean pooling and does not select CLS or max pooling. Keep that pooling recipe with the model when preparing comparable vectors.
Each returned representation has 384 components. An index or clustering stage works with those vectors using a chosen similarity configuration, while the underlying source record remains available for review. Keep the same model, input preparation and comparison settings across the items being compared. Another model returning 384 values is not, from that shape alone, the same embedding space.
The reference Sentence Transformer configuration uses a maximum sequence length of 128 tokens. Prepare a long source as short coherent units rather than one oversized paragraph. A sentence containing a rule may need the next sentence’s exception; retain that small unit together when it fits. A word or character count does not directly establish the token length.
For a set of short descriptions, preserve names and distinguishing properties which matter to the comparison. A record containing only a vague category can be difficult to separate from other records of that type. Store the source association outside the vector so a returned candidate can be checked against the original material.
In semantic search, represent the query and corpus material with this same model and return candidate records with their source text. In clustering, another stage forms groups from the vectors. Inspect the records near a cluster boundary before using the grouping as a product decision or an automatic label.
The useful output is a compact representation of short text linked to its original record. This exact paraphrase-MiniLM-L6-v2 checkpoint has its own pooling and sequence configuration. The all-MiniLM and multilingual paraphrase variants are separate model profiles; their training, language scope or input settings are not implied by the shared MiniLM name.
", "aboutText": "paraphrase-MiniLM-L6-v2 is a Sentence Transformers embedding model for sentences and short paragraphs. It maps each text into a 384-dimensional dense vector space, with semantic search and clustering among the maker’s stated tasks. The published architecture combines a transformer with mean pooling; its configured sequence length is 128 tokens.\n\n### Represent the passage you want to compare\n\nPrepare the actual sentence or short paragraph, keeping the subject and important conditions intact. A practical collection can contain differently worded questions or brief descriptions of related topics. Embed each record and retain its original wording and identifier so a later comparison can return the text itself.\n\nFor example, several help questions may describe the same difficulty with different terms. A similarity stage can bring those candidates together, after which their wording can be inspected for the actual issue. The embedding is the numerical representation used in that comparison; it does not write a paraphrase or assign a final cluster label on its own.\n\n### Mean pooling and vector use\n\nThe maker’s direct encoding example takes the attention mask into account when averaging token representations, excluding padding positions. Its architecture explicitly selects mean pooling and does not select CLS or max pooling. Keep that pooling recipe with the model when preparing comparable vectors.\n\nEach returned representation has 384 components. An index or clustering stage works with those vectors using a chosen similarity configuration, while the underlying source record remains available for review. Keep the same model, input preparation and comparison settings across the items being compared. Another model returning 384 values is not, from that shape alone, the same embedding space.\n\n### Short inputs in the published configuration\n\nThe reference Sentence Transformer configuration uses a maximum sequence length of 128 tokens. Prepare a long source as short coherent units rather than one oversized paragraph. A sentence containing a rule may need the next sentence’s exception; retain that small unit together when it fits. A word or character count does not directly establish the token length.\n\nFor a set of short descriptions, preserve names and distinguishing properties which matter to the comparison. A record containing only a vague category can be difficult to separate from other records of that type. Store the source association outside the vector so a returned candidate can be checked against the original material.\n\n### Search and clustering results\n\nIn semantic search, represent the query and corpus material with this same model and return candidate records with their source text. In clustering, another stage forms groups from the vectors. Inspect the records near a cluster boundary before using the grouping as a product decision or an automatic label.\n\nThe useful output is a compact representation of short text linked to its original record. This exact paraphrase-MiniLM-L6-v2 checkpoint has its own pooling and sequence configuration. The all-MiniLM and multilingual paraphrase variants are separate model profiles; their training, language scope or input settings are not implied by the shared MiniLM name.", "editorialOverlay": { "copy_fields": null, "copy_sha256": null, "copy_source": null, "html_sha256": "eaf735dc252df05f0d3607b996b8337912901d5e34f7c2f54be9389170950003", "module_sha256": "ee9dc1c50b1e6a8fb7fa062030fcb4646b9aaaeb08be513b723a5a92a7290f06", "raw_model_unchanged": true, "schema": "ailoot-native-model-editorial-v1" }, "locale": "en", "methods": [ { "html": "sentence-transformers/paraphrase-minilm-l6-v2Sentence Transformers paraphrase-MiniLM-L6-v2 maps sentences and short paragraphs into 384-dimensional vectors for semantic search and clustering. Its published configuration uses mean pooling and a 128-token sequence setting.
This description covers the model’s capabilities. Available in Ailoot inputs, outputs and options are listed for the selected method below.
paraphrase-MiniLM-L6-v2 is a Sentence Transformers embedding model for sentences and short paragraphs. It maps each text into a 384-dimensional dense vector space, with semantic search and clustering among the maker’s stated tasks. The published architecture combines a transformer with mean pooling; its configured sequence length is 128 tokens.
Prepare the actual sentence or short paragraph, keeping the subject and important conditions intact. A practical collection can contain differently worded questions or brief descriptions of related topics. Embed each record and retain its original wording and identifier so a later comparison can return the text itself.
For example, several help questions may describe the same difficulty with different terms. A similarity stage can bring those candidates together, after which their wording can be inspected for the actual issue. The embedding is the numerical representation used in that comparison; it does not write a paraphrase or assign a final cluster label on its own.
The maker’s direct encoding example takes the attention mask into account when averaging token representations, excluding padding positions. Its architecture explicitly selects mean pooling and does not select CLS or max pooling. Keep that pooling recipe with the model when preparing comparable vectors.
Each returned representation has 384 components. An index or clustering stage works with those vectors using a chosen similarity configuration, while the underlying source record remains available for review. Keep the same model, input preparation and comparison settings across the items being compared. Another model returning 384 values is not, from that shape alone, the same embedding space.
The reference Sentence Transformer configuration uses a maximum sequence length of 128 tokens. Prepare a long source as short coherent units rather than one oversized paragraph. A sentence containing a rule may need the next sentence’s exception; retain that small unit together when it fits. A word or character count does not directly establish the token length.
For a set of short descriptions, preserve names and distinguishing properties which matter to the comparison. A record containing only a vague category can be difficult to separate from other records of that type. Store the source association outside the vector so a returned candidate can be checked against the original material.
In semantic search, represent the query and corpus material with this same model and return candidate records with their source text. In clustering, another stage forms groups from the vectors. Inspect the records near a cluster boundary before using the grouping as a product decision or an automatic label.
The useful output is a compact representation of short text linked to its original record. This exact paraphrase-MiniLM-L6-v2 checkpoint has its own pooling and sequence configuration. The all-MiniLM and multilingual paraphrase variants are separate model profiles; their training, language scope or input settings are not implied by the shared MiniLM name.
POST /v1/embeddingstext, token_idsembedding| Field | Type | Requirement | Values & limits | Default |
|---|---|---|---|---|
Fieldmodel | Typestring | RequirementRequired | Values & limits sentence-transformers/paraphrase-minilm-l6-v2 | Default— |
Fieldinput | Typestring|string[]|integer[]|integer[][] | RequirementRequired | Values & limits Minimum items: 1 Conditions: Nonempty homogeneous batch. Token IDs are nonnegative integers. Input budget: Maximum: 512 · Text unit: UTF8 bytes · Token unit: token-array elements | Default— |
Fieldencoding_format | Typestring | RequirementOptional | Values & limits float | Defaultfloat |
Fielddimensions | Typeinteger | RequirementNot supported | Values & limits Minimum: 1 | Default— |
Fielduser | Typejson_value | RequirementOptional | Values & limits Conditions: Source allows this JSON metadata field; exact upstream semantics are model-specific. | Default— |
Fieldsession_id | Typejson_value | RequirementOptional | Values & limits Conditions: Source allows this JSON metadata field; exact upstream semantics are model-specific. | Default— |
Fieldtrace | Typejson_value | RequirementOptional | Values & limits Conditions: Source allows this JSON metadata field; exact upstream semantics are model-specific. | Default— |
Fieldprovider | Typeobject | RequirementOptional | Values & limits Conditions: provider.options rejected | Default— |
text_to_embedding
| Parameter | Contract |
|---|---|
Parametercontent_type | Contractapplication/json |
Parameterrequired | Contractmodel, input |
Parameterinput_shapes | Contractstring, string[], nonnegative_integer_token_array, nonnegative_integer_token_array[] |
Parameterencoding_format | Contractfloat |
Parameterdimensions_override | Contractfalse |
Parametermax_input_budget_units | Contract512 |
Parametermultimodal_objects | Contractfalse |
Parametermaterial_types | Contracttext, token_ids |
Lifecycle: synchronous_json
Keep every vector index and dimension.
Multimodal model name does not admit image objects on this exact path.
Tariff references do not add request options. Supported parameters are listed for the selected method.
| Configuration | Rate schedule | Charge |
|---|---|---|
| ConfigurationBase token rates | Rate scheduleBase rate | Charge Input · Input tokens $0.00001 / 1K tokens Output · Output tokens $0 / 1K tokens Input · Cache read — / 1K tokens Input · Cache write — / 1K tokens Input · Cache write, 1 hour — / 1K tokens |
Ailoot reserves an estimated amount before a request. After the result, it settles the final charge from measured usage. Streaming settlement uses the final usage record. How reservation and settlement work →
set -eu\n: "${AILOOT_API_KEY:?Set your Ailoot API key}"\npython3 - <<'PY_REQUEST'\nimport json,os\nfrom pathlib import Path\np=json.loads('{"model": "sentence-transformers/paraphrase-minilm-l6-v2", "input": ["First document", "Second document"]}')\n\nPath("request.json").write_text(json.dumps(p,ensure_ascii=False))\nPY_REQUEST\ncurl --fail-with-body --silent --show-error https://api.ailoot.ai/v1/embeddings \\\n -H "Authorization: Bearer $AILOOT_API_KEY" -H 'Content-Type: application/json' \\\n --data-binary @request.json -D response-headers.txt -o result.json\npython3 - <<'PY_RESULT'\nimport json\nfrom pathlib import Path\nr=json.loads(Path("result.json").read_text())\nfor x in r.get("data",[]): print(json.dumps({"index":x.get("index"),"embedding":x.get("embedding")}))\nPY_RESULT\nfrom openai import OpenAI\nimport os\n\nclient=OpenAI(base_url="https://api.ailoot.ai/v1",api_key=os.environ["AILOOT_API_KEY"],max_retries=0)\nresult=client.embeddings.create(model='sentence-transformers/paraphrase-minilm-l6-v2',input=["First document","Second document"])\nprint(result)\nfor item in result.data:print(item.index,item.embedding)\njson
Keep every complete returned response, file and content part.
Keep every vector with its index. Vector count must equal input count; dimensions must be consistent.
No embedding job endpoint is introduced.
| Field | Type |
|---|---|
Fieldobject | Typestring |
Fieldmodel | Typestring |
Fielddata[].index | Typeinteger |
Fielddata[].embedding | Typefinite float[] |
Fieldusage | Typeobject |
| Response | Condition | Next step |
|---|---|---|
Response400 · invalid_request_error | ConditionInvalid required fields, material representation or parameter value | Next stepCorrect the indicated fields before a new creation request. |
Ailoot Ultra sentence-transformers/paraphrase-minilm-l6-v2
Use the method paths shown above.