GTE-large is the large-sized General Text Embeddings model trained by Alibaba DAMO Academy. It is an English text encoder mainly based on the BERT framework, with a 1024-dimensional dense representation for each input. The maker trains the GTE family on relevance text pairs across domains and identifies retrieval, semantic textual similarity and text reranking as downstream applications.
Prepare the English passage that should be compared with a search query. For a collection of technical reports, retain a short section’s subject, observation and relevant condition. A report describing vibration only under a particular load should preserve that condition, rather than leaving a generic statement about vibration.
Keep source identifiers, document titles and section locations with the stored vectors. A query can retrieve candidate sections, while the original report remains available to verify the equipment, condition and conclusion. The embedding supplies a numerical representation of the passage’s content.
The published recipe uses average pooling over token representations with the attention mask, excluding padding from the average. It also shows optional L2 normalization followed by a dot-product comparison of normalized vectors. The Sentence Transformers example uses cosine similarity.
Retain the same preparation, pooling and comparison settings across the collection and the query side. Each representation contains 1024 components. The larger component count is an edition detail; it does not establish that GTE-large vectors can be mixed with the 768-dimensional GTE-base representation or another 1024-dimensional encoder.
A reranking workflow can compare a query embedding with an already retrieved set of candidate embeddings and reorder those records. Keep the distinction between that vector comparison and a separate model that jointly evaluates a query–passage pair.
The card explicitly restricts this exact model to English text and truncates long inputs to a maximum of 512 tokens. Split a long technical report into coherent sections and keep enough local context for each section to stand on its own. If a sentence depends on a preceding qualification, retain that qualification in the same indexed unit when it fits.
A token boundary is different from a visible word count. Verify the length of the prepared section and ensure that the detail needed by the intended search is not positioned beyond the part retained after truncation.
For a report-search application, present candidate sections with their source locations and inspect their actual content. Similar observations can have different causes or recommendations. A vector score helps order the candidates under the chosen comparison setup; it does not establish that a report’s conclusion applies to a new situation.
For RAG, use the selected sections as source context for the answer-writing stage and retain their document associations. This exact GTE-large profile supplies English dense embeddings with 1024 dimensions and a 512-token input boundary. Later GTE multilingual and extended-context editions have separate identities, so their language coverage or input settings do not follow from this checkpoint’s family name.
", "aboutText": "GTE-large is the large-sized General Text Embeddings model trained by Alibaba DAMO Academy. It is an English text encoder mainly based on the BERT framework, with a 1024-dimensional dense representation for each input. The maker trains the GTE family on relevance text pairs across domains and identifies retrieval, semantic textual similarity and text reranking as downstream applications.\n\n### Represent the content of a section\n\nPrepare the English passage that should be compared with a search query. For a collection of technical reports, retain a short section’s subject, observation and relevant condition. A report describing vibration only under a particular load should preserve that condition, rather than leaving a generic statement about vibration.\n\nKeep source identifiers, document titles and section locations with the stored vectors. A query can retrieve candidate sections, while the original report remains available to verify the equipment, condition and conclusion. The embedding supplies a numerical representation of the passage’s content.\n\n### Pooling, normalization and vector use\n\nThe published recipe uses average pooling over token representations with the attention mask, excluding padding from the average. It also shows optional L2 normalization followed by a dot-product comparison of normalized vectors. The Sentence Transformers example uses cosine similarity.\n\nRetain the same preparation, pooling and comparison settings across the collection and the query side. Each representation contains 1024 components. The larger component count is an edition detail; it does not establish that GTE-large vectors can be mixed with the 768-dimensional GTE-base representation or another 1024-dimensional encoder.\n\nA reranking workflow can compare a query embedding with an already retrieved set of candidate embeddings and reorder those records. Keep the distinction between that vector comparison and a separate model that jointly evaluates a query–passage pair.\n\n### Input length and English limitation\n\nThe card explicitly restricts this exact model to English text and truncates long inputs to a maximum of 512 tokens. Split a long technical report into coherent sections and keep enough local context for each section to stand on its own. If a sentence depends on a preceding qualification, retain that qualification in the same indexed unit when it fits.\n\nA token boundary is different from a visible word count. Verify the length of the prepared section and ensure that the detail needed by the intended search is not positioned beyond the part retained after truncation.\n\n### Inspect the retrieved evidence\n\nFor a report-search application, present candidate sections with their source locations and inspect their actual content. Similar observations can have different causes or recommendations. A vector score helps order the candidates under the chosen comparison setup; it does not establish that a report’s conclusion applies to a new situation.\n\nFor RAG, use the selected sections as source context for the answer-writing stage and retain their document associations. This exact GTE-large profile supplies English dense embeddings with 1024 dimensions and a 512-token input boundary. Later GTE multilingual and extended-context editions have separate identities, so their language coverage or input settings do not follow from this checkpoint’s family name.", "editorialOverlay": { "copy_fields": null, "copy_sha256": null, "copy_source": null, "html_sha256": "972e96e726f155d395d9cc50a6697f93d3781e84e169560c98294e13375e51ac", "module_sha256": "ee9dc1c50b1e6a8fb7fa062030fcb4646b9aaaeb08be513b723a5a92a7290f06", "raw_model_unchanged": true, "schema": "ailoot-native-model-editorial-v1" }, "locale": "en", "methods": [ { "html": "thenlper/gte-largeAlibaba DAMO Academy GTE-large converts English text into 1024-dimensional vectors for retrieval, semantic similarity and embedding-based reranking. Its reference recipe uses average pooling, optional normalization and a maximum sequence length of 512 tokens.
This description covers the model’s capabilities. Available in Ailoot inputs, outputs and options are listed for the selected method below.
GTE-large is the large-sized General Text Embeddings model trained by Alibaba DAMO Academy. It is an English text encoder mainly based on the BERT framework, with a 1024-dimensional dense representation for each input. The maker trains the GTE family on relevance text pairs across domains and identifies retrieval, semantic textual similarity and text reranking as downstream applications.
Prepare the English passage that should be compared with a search query. For a collection of technical reports, retain a short section’s subject, observation and relevant condition. A report describing vibration only under a particular load should preserve that condition, rather than leaving a generic statement about vibration.
Keep source identifiers, document titles and section locations with the stored vectors. A query can retrieve candidate sections, while the original report remains available to verify the equipment, condition and conclusion. The embedding supplies a numerical representation of the passage’s content.
The published recipe uses average pooling over token representations with the attention mask, excluding padding from the average. It also shows optional L2 normalization followed by a dot-product comparison of normalized vectors. The Sentence Transformers example uses cosine similarity.
Retain the same preparation, pooling and comparison settings across the collection and the query side. Each representation contains 1024 components. The larger component count is an edition detail; it does not establish that GTE-large vectors can be mixed with the 768-dimensional GTE-base representation or another 1024-dimensional encoder.
A reranking workflow can compare a query embedding with an already retrieved set of candidate embeddings and reorder those records. Keep the distinction between that vector comparison and a separate model that jointly evaluates a query–passage pair.
The card explicitly restricts this exact model to English text and truncates long inputs to a maximum of 512 tokens. Split a long technical report into coherent sections and keep enough local context for each section to stand on its own. If a sentence depends on a preceding qualification, retain that qualification in the same indexed unit when it fits.
A token boundary is different from a visible word count. Verify the length of the prepared section and ensure that the detail needed by the intended search is not positioned beyond the part retained after truncation.
For a report-search application, present candidate sections with their source locations and inspect their actual content. Similar observations can have different causes or recommendations. A vector score helps order the candidates under the chosen comparison setup; it does not establish that a report’s conclusion applies to a new situation.
For RAG, use the selected sections as source context for the answer-writing stage and retain their document associations. This exact GTE-large profile supplies English dense embeddings with 1024 dimensions and a 512-token input boundary. Later GTE multilingual and extended-context editions have separate identities, so their language coverage or input settings do not follow from this checkpoint’s family name.
POST /v1/embeddingstext, token_idsembedding| Field | Type | Requirement | Values & limits | Default |
|---|---|---|---|---|
Fieldmodel | Typestring | RequirementRequired | Values & limits thenlper/gte-large | 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 $0.00002 / 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": "thenlper/gte-large", "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='thenlper/gte-large',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 thenlper/gte-large
Use the method paths shown above.