Voyage Finance 2 is a text embedding model specialized for retrieval over financial material. It converts supplied passages and search questions into vectors that help an application locate relevant source text. The maker’s finance evaluation includes public filings, financial reports, corporate news and finance question-answering material, reflecting the edition’s focus on finding domain-relevant evidence.
Prepare the corpus so financial meaning survives extraction. Keep a figure with its label, period, unit and surrounding explanation. A revenue value without the company name or reporting period is difficult to interpret correctly in a search result. For a table converted to text, retain row and column headings and identify which values belong together. Preserve the original document and page or section location for review.
The model supports a 32,000-token context window and returns 1,024-dimensional embeddings. Use the available input space for complete, understandable passages while selecting boundaries that suit the intended result. A single broad section may help locate a report; a focused passage can show the specific statement that answers a question. Keep both the document relationship and the passage location in the index metadata.
Write retrieval questions with the distinctions that matter to the task. “What explains the change in operating margin for the latest reported quarter?” requires a different passage from a question asking only for the margin value. Include the company, reporting period or business segment when it is known. Test similar questions against documents with nearby dates and comparable terminology to see whether the correct evidence is returned.
Use the same model for embedding the corpus and the search queries. Keep every output vector mapped to its input passage, then use the retrieval result to show the original text and source reference. A finance assistant can pass selected evidence to an answer-generation step while retaining the citations needed to check any stated amount or explanation.
Separate reports, news and explanatory material through metadata when the application needs those distinctions. Store publication date and reporting period independently; they describe different aspects of a financial record. When a document is revised, preserve the version relationship and update the affected passages so a search result leads to the intended source.
Evaluate the system with questions whose answers depend on financial terminology and document context. Useful cases include a passage explaining a change, a table containing the required value and several reports that use the same measure for different periods. Inspect whether the returned text contains enough information to identify the relevant entity and period without guessing from the query.
The model supplies retrieval vectors for a financial knowledge base, research archive or evidence-selection workflow. Keep numerical verification and the reading of the underlying document in the application’s review process. A relevant passage is a starting point for checking a value, relationship or claim; its source content remains the basis for the final explanation. Retain a real question set with expected passages so improvements to extraction and indexing can be assessed on the same financial tasks.
", "aboutText": "Voyage Finance 2 is a text embedding model specialized for retrieval over financial material. It converts supplied passages and search questions into vectors that help an application locate relevant source text. The maker’s finance evaluation includes public filings, financial reports, corporate news and finance question-answering material, reflecting the edition’s focus on finding domain-relevant evidence.\n\nPrepare the corpus so financial meaning survives extraction. Keep a figure with its label, period, unit and surrounding explanation. A revenue value without the company name or reporting period is difficult to interpret correctly in a search result. For a table converted to text, retain row and column headings and identify which values belong together. Preserve the original document and page or section location for review.\n\nThe model supports a 32,000-token context window and returns 1,024-dimensional embeddings. Use the available input space for complete, understandable passages while selecting boundaries that suit the intended result. A single broad section may help locate a report; a focused passage can show the specific statement that answers a question. Keep both the document relationship and the passage location in the index metadata.\n\nWrite retrieval questions with the distinctions that matter to the task. “What explains the change in operating margin for the latest reported quarter?” requires a different passage from a question asking only for the margin value. Include the company, reporting period or business segment when it is known. Test similar questions against documents with nearby dates and comparable terminology to see whether the correct evidence is returned.\n\nUse the same model for embedding the corpus and the search queries. Keep every output vector mapped to its input passage, then use the retrieval result to show the original text and source reference. A finance assistant can pass selected evidence to an answer-generation step while retaining the citations needed to check any stated amount or explanation.\n\nSeparate reports, news and explanatory material through metadata when the application needs those distinctions. Store publication date and reporting period independently; they describe different aspects of a financial record. When a document is revised, preserve the version relationship and update the affected passages so a search result leads to the intended source.\n\nEvaluate the system with questions whose answers depend on financial terminology and document context. Useful cases include a passage explaining a change, a table containing the required value and several reports that use the same measure for different periods. Inspect whether the returned text contains enough information to identify the relevant entity and period without guessing from the query.\n\nThe model supplies retrieval vectors for a financial knowledge base, research archive or evidence-selection workflow. Keep numerical verification and the reading of the underlying document in the application’s review process. A relevant passage is a starting point for checking a value, relationship or claim; its source content remains the basis for the final explanation. Retain a real question set with expected passages so improvements to extraction and indexing can be assessed on the same financial tasks.", "editorialOverlay": { "copy_fields": null, "copy_sha256": null, "copy_source": null, "html_sha256": "38764c1f54c8281390e16809c48a3c539e9fe6e49da579d88f5d745d12e1830c", "module_sha256": "ee9dc1c50b1e6a8fb7fa062030fcb4646b9aaaeb08be513b723a5a92a7290f06", "raw_model_unchanged": true, "schema": "ailoot-native-model-editorial-v1" }, "locale": "en", "methods": [ { "html": "voyage-finance-2Voyage Finance 2 embeds financial text for document retrieval and RAG. Its 32,000-token window and 1,024-dimensional vectors support search across filings, financial reports and finance news while keeping results linked to the original evidence.
This description covers the model’s capabilities. Available in Ailoot inputs, outputs and options are listed for the selected method below.
Voyage Finance 2 is a text embedding model specialized for retrieval over financial material. It converts supplied passages and search questions into vectors that help an application locate relevant source text. The maker’s finance evaluation includes public filings, financial reports, corporate news and finance question-answering material, reflecting the edition’s focus on finding domain-relevant evidence.
Prepare the corpus so financial meaning survives extraction. Keep a figure with its label, period, unit and surrounding explanation. A revenue value without the company name or reporting period is difficult to interpret correctly in a search result. For a table converted to text, retain row and column headings and identify which values belong together. Preserve the original document and page or section location for review.
The model supports a 32,000-token context window and returns 1,024-dimensional embeddings. Use the available input space for complete, understandable passages while selecting boundaries that suit the intended result. A single broad section may help locate a report; a focused passage can show the specific statement that answers a question. Keep both the document relationship and the passage location in the index metadata.
Write retrieval questions with the distinctions that matter to the task. “What explains the change in operating margin for the latest reported quarter?” requires a different passage from a question asking only for the margin value. Include the company, reporting period or business segment when it is known. Test similar questions against documents with nearby dates and comparable terminology to see whether the correct evidence is returned.
Use the same model for embedding the corpus and the search queries. Keep every output vector mapped to its input passage, then use the retrieval result to show the original text and source reference. A finance assistant can pass selected evidence to an answer-generation step while retaining the citations needed to check any stated amount or explanation.
Separate reports, news and explanatory material through metadata when the application needs those distinctions. Store publication date and reporting period independently; they describe different aspects of a financial record. When a document is revised, preserve the version relationship and update the affected passages so a search result leads to the intended source.
Evaluate the system with questions whose answers depend on financial terminology and document context. Useful cases include a passage explaining a change, a table containing the required value and several reports that use the same measure for different periods. Inspect whether the returned text contains enough information to identify the relevant entity and period without guessing from the query.
The model supplies retrieval vectors for a financial knowledge base, research archive or evidence-selection workflow. Keep numerical verification and the reading of the underlying document in the application’s review process. A relevant passage is a starting point for checking a value, relationship or claim; its source content remains the basis for the final explanation. Retain a real question set with expected passages so improvements to extraction and indexing can be assessed on the same financial tasks.
POST /v1/embeddingstextembedding| Field | Type | Requirement | Values & limits | Default |
|---|---|---|---|---|
Fieldmodel | Typestring | RequirementRequired | Values & limits voyage-finance-2 Conditions: Use the exact Ailoot model selector. | Default— |
Fieldinput | Typestring or array<string> | RequirementRequired | Values & limits Conditions: A text or list of texts; keep one vector for every returned data item and its index. | Default— |
| Parameter | Contract |
|---|---|
Parametercontent_type | Contractapplication/json |
Parameterrequired | Contractmodel, input |
Parameteroptional | Contract |
Parametermaterial_types | Contracttext |
Parameterconditions | ContractSend text with the selected model; retain every returned vector and its index. |
Lifecycle: synchronous
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 $0.0004 / 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}"\ncat > request.json <<'JSON'\n{\n "model": "voyage-finance-2",\n "input": "The quick brown fox jumps over the lazy dog"\n}\nJSON\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 embeddings.json\n# Keep every data item and its index, the full usage object and response headers.\nAiloot Ultra voyage-finance-2
Use the method paths shown above.