all-mpnet-base-v2 is a Sentence Transformers model for representing sentences and short paragraphs as 768-dimensional dense vectors. It was fine-tuned from the MPNet base model with contrastive learning on a billion-pair sentence collection. The maker identifies information retrieval, clustering and sentence similarity as intended uses.
Prepare the text that should be compared, keeping the main idea and its distinguishing details. For an internal document collection, a short section might describe how to request an equipment replacement or what information a maintenance report must contain. Keep its document title and source location available alongside the vector.
A query embedded with the same model can retrieve related sections even when their wording differs. Inspect those sections for the actual instruction needed by the reader. For a similarity workflow, compare the candidate sentences themselves after retrieval, especially where one includes an exception or condition absent from the other.
The reference encoding example averages contextual token representations with the attention mask, excluding padding from the mean. It then L2-normalizes the resulting sentence embeddings. Preserve both steps when producing vectors for a collection and for later queries.
Each vector has 768 components. Cosine similarity can compare those normalized representations, while the source text provides the evidence behind a result. The shared component count does not make this model compatible with multi-qa-mpnet-base-dot-v1: that separate edition specifies CLS pooling and unnormalized vectors for dot-product scoring.
The maker states that input beyond 384 word pieces is truncated by default. Keep the relevant section within that tokenizer boundary. For a long procedure, split at meaningful section boundaries and include enough local context to explain what a step applies to. A fragment such as “Submit within five days” needs the event or process that starts the deadline.
The training procedure limited sequences to 128 tokens, which is separate from the published 384-word-piece encoding default. Use representative material when assessing longer paragraphs or vocabulary specific to your collection. More text fitting into the input does not establish that every detail will influence a retrieval result equally.
For search, keep the vector-to-document association so ranked candidates can return the original sections. In a RAG workflow, selected sections become context for the answer-writing stage. The model supplies numerical representations; it does not generate a sourced response or decide which document is authoritative.
For clustering, compare the records in each proposed group before assigning labels. A set of maintenance notes may contain similar language but refer to different devices or failure conditions. Include those distinguishing details in the encoded text and inspect them in the returned material rather than relying on a score alone.
This exact all-mpnet-base-v2 checkpoint is an English sentence-embedding profile with its own mean-pooling, normalization and default truncation settings. Keep those settings with the indexed collection and use the vectors as an aid to finding and comparing the source passages that matter for the task.
", "aboutText": "all-mpnet-base-v2 is a Sentence Transformers model for representing sentences and short paragraphs as 768-dimensional dense vectors. It was fine-tuned from the MPNet base model with contrastive learning on a billion-pair sentence collection. The maker identifies information retrieval, clustering and sentence similarity as intended uses.\n\n### Encode a sentence or a short paragraph\n\nPrepare the text that should be compared, keeping the main idea and its distinguishing details. For an internal document collection, a short section might describe how to request an equipment replacement or what information a maintenance report must contain. Keep its document title and source location available alongside the vector.\n\nA query embedded with the same model can retrieve related sections even when their wording differs. Inspect those sections for the actual instruction needed by the reader. For a similarity workflow, compare the candidate sentences themselves after retrieval, especially where one includes an exception or condition absent from the other.\n\n### Mean pooling and normalization\n\nThe reference encoding example averages contextual token representations with the attention mask, excluding padding from the mean. It then L2-normalizes the resulting sentence embeddings. Preserve both steps when producing vectors for a collection and for later queries.\n\nEach vector has 768 components. Cosine similarity can compare those normalized representations, while the source text provides the evidence behind a result. The shared component count does not make this model compatible with multi-qa-mpnet-base-dot-v1: that separate edition specifies CLS pooling and unnormalized vectors for dot-product scoring.\n\n### Default truncation and passage preparation\n\nThe maker states that input beyond 384 word pieces is truncated by default. Keep the relevant section within that tokenizer boundary. For a long procedure, split at meaningful section boundaries and include enough local context to explain what a step applies to. A fragment such as “Submit within five days” needs the event or process that starts the deadline.\n\nThe training procedure limited sequences to 128 tokens, which is separate from the published 384-word-piece encoding default. Use representative material when assessing longer paragraphs or vocabulary specific to your collection. More text fitting into the input does not establish that every detail will influence a retrieval result equally.\n\n### Retrieval, grouping and comparison\n\nFor search, keep the vector-to-document association so ranked candidates can return the original sections. In a RAG workflow, selected sections become context for the answer-writing stage. The model supplies numerical representations; it does not generate a sourced response or decide which document is authoritative.\n\nFor clustering, compare the records in each proposed group before assigning labels. A set of maintenance notes may contain similar language but refer to different devices or failure conditions. Include those distinguishing details in the encoded text and inspect them in the returned material rather than relying on a score alone.\n\nThis exact all-mpnet-base-v2 checkpoint is an English sentence-embedding profile with its own mean-pooling, normalization and default truncation settings. Keep those settings with the indexed collection and use the vectors as an aid to finding and comparing the source passages that matter for the task.", "editorialOverlay": { "copy_fields": null, "copy_sha256": null, "copy_source": null, "html_sha256": "ab4d20838e10fea35de76a3fcb74e8f3869d0c1cbd3f0784d74a9c1f1bd84d37", "module_sha256": "ee9dc1c50b1e6a8fb7fa062030fcb4646b9aaaeb08be513b723a5a92a7290f06", "raw_model_unchanged": true, "schema": "ailoot-native-model-editorial-v1" }, "locale": "en", "methods": [ { "html": "sentence-transformers/all-mpnet-base-v2Sentence Transformers all-mpnet-base-v2 converts sentences and short paragraphs into 768-dimensional vectors for retrieval, clustering and similarity. The published recipe uses mean pooling and L2 normalization, with default truncation at 384 word pieces.
This description covers the model’s capabilities. Available in Ailoot inputs, outputs and options are listed for the selected method below.
all-mpnet-base-v2 is a Sentence Transformers model for representing sentences and short paragraphs as 768-dimensional dense vectors. It was fine-tuned from the MPNet base model with contrastive learning on a billion-pair sentence collection. The maker identifies information retrieval, clustering and sentence similarity as intended uses.
Prepare the text that should be compared, keeping the main idea and its distinguishing details. For an internal document collection, a short section might describe how to request an equipment replacement or what information a maintenance report must contain. Keep its document title and source location available alongside the vector.
A query embedded with the same model can retrieve related sections even when their wording differs. Inspect those sections for the actual instruction needed by the reader. For a similarity workflow, compare the candidate sentences themselves after retrieval, especially where one includes an exception or condition absent from the other.
The reference encoding example averages contextual token representations with the attention mask, excluding padding from the mean. It then L2-normalizes the resulting sentence embeddings. Preserve both steps when producing vectors for a collection and for later queries.
Each vector has 768 components. Cosine similarity can compare those normalized representations, while the source text provides the evidence behind a result. The shared component count does not make this model compatible with multi-qa-mpnet-base-dot-v1: that separate edition specifies CLS pooling and unnormalized vectors for dot-product scoring.
The maker states that input beyond 384 word pieces is truncated by default. Keep the relevant section within that tokenizer boundary. For a long procedure, split at meaningful section boundaries and include enough local context to explain what a step applies to. A fragment such as “Submit within five days” needs the event or process that starts the deadline.
The training procedure limited sequences to 128 tokens, which is separate from the published 384-word-piece encoding default. Use representative material when assessing longer paragraphs or vocabulary specific to your collection. More text fitting into the input does not establish that every detail will influence a retrieval result equally.
For search, keep the vector-to-document association so ranked candidates can return the original sections. In a RAG workflow, selected sections become context for the answer-writing stage. The model supplies numerical representations; it does not generate a sourced response or decide which document is authoritative.
For clustering, compare the records in each proposed group before assigning labels. A set of maintenance notes may contain similar language but refer to different devices or failure conditions. Include those distinguishing details in the encoded text and inspect them in the returned material rather than relying on a score alone.
This exact all-mpnet-base-v2 checkpoint is an English sentence-embedding profile with its own mean-pooling, normalization and default truncation settings. Keep those settings with the indexed collection and use the vectors as an aid to finding and comparing the source passages that matter for the task.
POST /v1/embeddingstext, token_idsembedding| Field | Type | Requirement | Values & limits | Default |
|---|---|---|---|---|
Fieldmodel | Typestring | RequirementRequired | Values & limits sentence-transformers/all-mpnet-base-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/all-mpnet-base-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/all-mpnet-base-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/all-mpnet-base-v2
Use the method paths shown above.