GTE-base is the base-sized General Text Embeddings model trained by Alibaba DAMO Academy. This English text encoder is mainly based on the BERT framework and produces 768-dimensional dense vectors. Its training uses a large corpus of relevance pairs across domains, with information retrieval, semantic textual similarity and embedding-based text reranking among the stated applications.
Use readable English text for both the information need and the material to be searched. For example, a query about replacing a worn component can be compared with short maintenance sections describing replacement criteria, inspection steps and ordering instructions. Retain the component name and the conditions that distinguish those sections.
The published example encodes questions and candidate text directly, without an added task instruction. Keep that preparation consistent across the collection and later queries. Store the original passage and its identifier alongside the vector so a ranked result can show the information that was represented.
The reference recipe averages the model’s token representations using the attention mask, preventing padding positions from contributing to the result. It then shows optional L2 normalization and compares the normalized vectors through their dot product. A separate example uses cosine similarity.
Choose and retain a consistent normalization and comparison setup for indexing and retrieval. The output contains 768 numerical components; it does not contain a generated answer. For a candidate-reranking stage, encode the candidate passages with the same model and order them by their similarity to the query. This remains embedding-based comparison, rather than a claim that the model jointly reads every query–passage pair as a cross-encoder.
The exact card limits this model to English text and states that long inputs are truncated to a maximum of 512 tokens. A long manual should therefore become meaningful sections before indexing. Keep a warning, exception or qualification with the rule it modifies whenever those details are needed to interpret the passage.
For a query about maintenance intervals, a section containing the interval and the operating condition is more useful than a fragment containing only a number. Preserve the heading or equipment identity when it disambiguates the instruction. Check tokenizer length rather than relying on the visible number of words.
Retrieval returns candidates from the indexed collection. Inspect the leading passages for the exact task, and retain their source locations when passing them into a RAG answer-writing stage. A related passage may discuss inspection without authorizing replacement; that distinction must remain visible in the source material.
GTE-base is distinct from GTE-large, which produces 1024-dimensional vectors, and from later multilingual or extended-context GTE editions. Its published model identity defines the English scope, 768-component representation and 512-token boundary. Use this exact profile with representative queries and source records when assessing a retrieval collection.
", "aboutText": "GTE-base is the base-sized General Text Embeddings model trained by Alibaba DAMO Academy. This English text encoder is mainly based on the BERT framework and produces 768-dimensional dense vectors. Its training uses a large corpus of relevance pairs across domains, with information retrieval, semantic textual similarity and embedding-based text reranking among the stated applications.\n\n### Prepare English queries and passages\n\nUse readable English text for both the information need and the material to be searched. For example, a query about replacing a worn component can be compared with short maintenance sections describing replacement criteria, inspection steps and ordering instructions. Retain the component name and the conditions that distinguish those sections.\n\nThe published example encodes questions and candidate text directly, without an added task instruction. Keep that preparation consistent across the collection and later queries. Store the original passage and its identifier alongside the vector so a ranked result can show the information that was represented.\n\n### Average pooling and comparison\n\nThe reference recipe averages the model’s token representations using the attention mask, preventing padding positions from contributing to the result. It then shows optional L2 normalization and compares the normalized vectors through their dot product. A separate example uses cosine similarity.\n\nChoose and retain a consistent normalization and comparison setup for indexing and retrieval. The output contains 768 numerical components; it does not contain a generated answer. For a candidate-reranking stage, encode the candidate passages with the same model and order them by their similarity to the query. This remains embedding-based comparison, rather than a claim that the model jointly reads every query–passage pair as a cross-encoder.\n\n### English scope and 512-token inputs\n\nThe exact card limits this model to English text and states that long inputs are truncated to a maximum of 512 tokens. A long manual should therefore become meaningful sections before indexing. Keep a warning, exception or qualification with the rule it modifies whenever those details are needed to interpret the passage.\n\nFor a query about maintenance intervals, a section containing the interval and the operating condition is more useful than a fragment containing only a number. Preserve the heading or equipment identity when it disambiguates the instruction. Check tokenizer length rather than relying on the visible number of words.\n\n### Use source passages behind the score\n\nRetrieval returns candidates from the indexed collection. Inspect the leading passages for the exact task, and retain their source locations when passing them into a RAG answer-writing stage. A related passage may discuss inspection without authorizing replacement; that distinction must remain visible in the source material.\n\nGTE-base is distinct from GTE-large, which produces 1024-dimensional vectors, and from later multilingual or extended-context GTE editions. Its published model identity defines the English scope, 768-component representation and 512-token boundary. Use this exact profile with representative queries and source records when assessing a retrieval collection.", "editorialOverlay": { "copy_fields": null, "copy_sha256": null, "copy_source": null, "html_sha256": "5cade9852c3d8ccfa72c81c3fa140e006958aea4fa0e861b13e4437d98de4165", "module_sha256": "ee9dc1c50b1e6a8fb7fa062030fcb4646b9aaaeb08be513b723a5a92a7290f06", "raw_model_unchanged": true, "schema": "ailoot-native-model-editorial-v1" }, "locale": "en", "methods": [ { "html": "thenlper/gte-baseAlibaba DAMO Academy GTE-base embeds English text into 768-dimensional vectors for retrieval and semantic similarity. The exact model uses average pooling, supports sequences up to 512 tokens and documents cosine-style comparison with optional normalization.
This description covers the model’s capabilities. Available in Ailoot inputs, outputs and options are listed for the selected method below.
GTE-base is the base-sized General Text Embeddings model trained by Alibaba DAMO Academy. This English text encoder is mainly based on the BERT framework and produces 768-dimensional dense vectors. Its training uses a large corpus of relevance pairs across domains, with information retrieval, semantic textual similarity and embedding-based text reranking among the stated applications.
Use readable English text for both the information need and the material to be searched. For example, a query about replacing a worn component can be compared with short maintenance sections describing replacement criteria, inspection steps and ordering instructions. Retain the component name and the conditions that distinguish those sections.
The published example encodes questions and candidate text directly, without an added task instruction. Keep that preparation consistent across the collection and later queries. Store the original passage and its identifier alongside the vector so a ranked result can show the information that was represented.
The reference recipe averages the model’s token representations using the attention mask, preventing padding positions from contributing to the result. It then shows optional L2 normalization and compares the normalized vectors through their dot product. A separate example uses cosine similarity.
Choose and retain a consistent normalization and comparison setup for indexing and retrieval. The output contains 768 numerical components; it does not contain a generated answer. For a candidate-reranking stage, encode the candidate passages with the same model and order them by their similarity to the query. This remains embedding-based comparison, rather than a claim that the model jointly reads every query–passage pair as a cross-encoder.
The exact card limits this model to English text and states that long inputs are truncated to a maximum of 512 tokens. A long manual should therefore become meaningful sections before indexing. Keep a warning, exception or qualification with the rule it modifies whenever those details are needed to interpret the passage.
For a query about maintenance intervals, a section containing the interval and the operating condition is more useful than a fragment containing only a number. Preserve the heading or equipment identity when it disambiguates the instruction. Check tokenizer length rather than relying on the visible number of words.
Retrieval returns candidates from the indexed collection. Inspect the leading passages for the exact task, and retain their source locations when passing them into a RAG answer-writing stage. A related passage may discuss inspection without authorizing replacement; that distinction must remain visible in the source material.
GTE-base is distinct from GTE-large, which produces 1024-dimensional vectors, and from later multilingual or extended-context GTE editions. Its published model identity defines the English scope, 768-component representation and 512-token boundary. Use this exact profile with representative queries and source records when assessing a retrieval collection.
POST /v1/embeddingstext, token_idsembedding| Field | Type | Requirement | Values & limits | Default |
|---|---|---|---|---|
Fieldmodel | Typestring | RequirementRequired | Values & limits thenlper/gte-base | 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": "thenlper/gte-base", "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-base',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-base
Use the method paths shown above.