Mistral Embed 2312 is Mistral AI's text embedding model for representing meaning numerically. It accepts a text string or a list of strings and returns an embedding for each input. The result is a vector used by the application for similarity, retrieval, classification or clustering rather than a conversational explanation. The maker's fixed model page identifies mistral-embed-2312.
Input, context and vector shape
The exact model page documents an 8k context label. Prepare each text item within the applicable input capacity and keep the item's identity with its vector. A long source document can be divided into useful passages while retaining the document title, section and source location needed to return the correct material later.
The maker's Mistral Embed guide specifies 1,024 dimensions for each text embedding, regardless of the input's visible length. A short question and a longer passage therefore have the same vector shape even though their contents and token counts differ. This is an output representation dimension, not an answer-token budget or a measure of how many source words are retained.
The documented response includes the embeddings and token-usage information; the endpoint example associates each vector with an input index. Preserve that association when processing several texts. The application needs the original text and its identifiers as well as the numerical result to show a useful search hit or provide source material to a later step.
Semantic similarity
The guide explains that related texts tend to be close in the embedding space. This supports finding passages that use different words to express the same topic or requirement. An illustrative knowledge-base query might ask how to recover access, while the relevant procedure uses a different phrase for the same operation. The application can compare the query's vector with stored passage vectors instead of relying only on exact word matching.
The maker's examples use unit-normalized embeddings. For such vectors, cosine similarity, dot product and Euclidean distance give consistent similarity rankings, even though their numerical values are not identical. Choose one suitable comparison method and apply it consistently across the query and document representations.
Similarity is not proof that two statements are interchangeable. Retain the source text and inspect the condition, date or identifier that matters to the question. Two passages can concern the same subject while describing different procedures or exceptions. The embedding provides a retrieval signal; interpreting the selected passage is a separate part of the task.
Retrieval and source-based answers
A practical document workflow prepares vectors for knowledge-base passages, stores their identifiers and original text, and then compares a new question with that collection. Return the relevant passages with their source locations. The application can show those passages directly or supply them to a text-generation model that writes an answer grounded in the retrieved material.
For a useful search result, retain the document title, section, passage and the identifier used to find it. A support assistant can then retrieve the procedure relevant to a user's question and pass that procedure to the answer-writing stage. Mistral Embed performs the numerical representation step; it does not itself draft the final explanation or execute an external action.
Other documented uses
The maker also lists paraphrase detection, classification, clustering and batched processing among embedding workflows. A comparison task can retrieve passages with related meaning for review. A classification workflow can compare an input with labeled examples or use the vectors as features for a separate classifier. A clustering workflow can group similar items for a later human or application-level description.
Define the purpose of the representation before preparing the data. A passage used for semantic search should preserve the information someone will need after retrieval; a classification example should include the wording that distinguishes the class. Keep the preparation consistent so that a query and a stored item represent comparable material.
Edition and output boundaries
Use the same model edition and compatible representation for the documents and their queries. A vector with the same number of coordinates from a different model is not automatically in the same semantic space. Replacing the model or representation requires the stored items and queries to remain aligned.
This entry describes text input and numerical vector output, with the maker's 8k context and Mistral Embed's documented 1,024-dimensional representation. The configurable dimensions and quantized formats of Codestral Embed belong to that separately named code model and should not be assumed here. Images, audio, generated prose and tool execution are separate capabilities outside this text-embedding result.
", "aboutText": "Mistral Embed 2312 is Mistral AI's text embedding model for representing meaning numerically. It accepts a text string or a list of strings and returns an embedding for each input. The result is a vector used by the application for similarity, retrieval, classification or clustering rather than a conversational explanation. The maker's fixed model page identifies mistral-embed-2312.\n\nInput, context and vector shape\n\nThe exact model page documents an 8k context label. Prepare each text item within the applicable input capacity and keep the item's identity with its vector. A long source document can be divided into useful passages while retaining the document title, section and source location needed to return the correct material later.\n\nThe maker's Mistral Embed guide specifies 1,024 dimensions for each text embedding, regardless of the input's visible length. A short question and a longer passage therefore have the same vector shape even though their contents and token counts differ. This is an output representation dimension, not an answer-token budget or a measure of how many source words are retained.\n\nThe documented response includes the embeddings and token-usage information; the endpoint example associates each vector with an input index. Preserve that association when processing several texts. The application needs the original text and its identifiers as well as the numerical result to show a useful search hit or provide source material to a later step.\n\nSemantic similarity\n\nThe guide explains that related texts tend to be close in the embedding space. This supports finding passages that use different words to express the same topic or requirement. An illustrative knowledge-base query might ask how to recover access, while the relevant procedure uses a different phrase for the same operation. The application can compare the query's vector with stored passage vectors instead of relying only on exact word matching.\n\nThe maker's examples use unit-normalized embeddings. For such vectors, cosine similarity, dot product and Euclidean distance give consistent similarity rankings, even though their numerical values are not identical. Choose one suitable comparison method and apply it consistently across the query and document representations.\n\nSimilarity is not proof that two statements are interchangeable. Retain the source text and inspect the condition, date or identifier that matters to the question. Two passages can concern the same subject while describing different procedures or exceptions. The embedding provides a retrieval signal; interpreting the selected passage is a separate part of the task.\n\nRetrieval and source-based answers\n\nA practical document workflow prepares vectors for knowledge-base passages, stores their identifiers and original text, and then compares a new question with that collection. Return the relevant passages with their source locations. The application can show those passages directly or supply them to a text-generation model that writes an answer grounded in the retrieved material.\n\nFor a useful search result, retain the document title, section, passage and the identifier used to find it. A support assistant can then retrieve the procedure relevant to a user's question and pass that procedure to the answer-writing stage. Mistral Embed performs the numerical representation step; it does not itself draft the final explanation or execute an external action.\n\nOther documented uses\n\nThe maker also lists paraphrase detection, classification, clustering and batched processing among embedding workflows. A comparison task can retrieve passages with related meaning for review. A classification workflow can compare an input with labeled examples or use the vectors as features for a separate classifier. A clustering workflow can group similar items for a later human or application-level description.\n\nDefine the purpose of the representation before preparing the data. A passage used for semantic search should preserve the information someone will need after retrieval; a classification example should include the wording that distinguishes the class. Keep the preparation consistent so that a query and a stored item represent comparable material.\n\nEdition and output boundaries\n\nUse the same model edition and compatible representation for the documents and their queries. A vector with the same number of coordinates from a different model is not automatically in the same semantic space. Replacing the model or representation requires the stored items and queries to remain aligned.\n\nThis entry describes text input and numerical vector output, with the maker's 8k context and Mistral Embed's documented 1,024-dimensional representation. The configurable dimensions and quantized formats of Codestral Embed belong to that separately named code model and should not be assumed here. Images, audio, generated prose and tool execution are separate capabilities outside this text-embedding result.", "editorialOverlay": { "copy_fields": null, "copy_sha256": null, "copy_source": null, "html_sha256": "00e76fbe298869e645d60676c56e4fe0e59efbb3c8e8c2785f8e4bbe50a44c05", "module_sha256": "ee9dc1c50b1e6a8fb7fa062030fcb4646b9aaaeb08be513b723a5a92a7290f06", "raw_model_unchanged": true, "schema": "ailoot-native-model-editorial-v1" }, "locale": "en", "methods": [ { "html": "mistralai/mistral-embed-2312Mistral Embed 2312 turns text into 1,024-dimensional vectors for semantic search, retrieval and related-text analysis. Its maker page documents an 8k context; the result is a numerical representation linked to each input, rather than a generated answer.
This description covers the model’s capabilities. Available in Ailoot inputs, outputs and options are listed for the selected method below.
Mistral Embed 2312 is Mistral AI's text embedding model for representing meaning numerically. It accepts a text string or a list of strings and returns an embedding for each input. The result is a vector used by the application for similarity, retrieval, classification or clustering rather than a conversational explanation. The maker's fixed model page identifies mistral-embed-2312.
Input, context and vector shape
The exact model page documents an 8k context label. Prepare each text item within the applicable input capacity and keep the item's identity with its vector. A long source document can be divided into useful passages while retaining the document title, section and source location needed to return the correct material later.
The maker's Mistral Embed guide specifies 1,024 dimensions for each text embedding, regardless of the input's visible length. A short question and a longer passage therefore have the same vector shape even though their contents and token counts differ. This is an output representation dimension, not an answer-token budget or a measure of how many source words are retained.
The documented response includes the embeddings and token-usage information; the endpoint example associates each vector with an input index. Preserve that association when processing several texts. The application needs the original text and its identifiers as well as the numerical result to show a useful search hit or provide source material to a later step.
Semantic similarity
The guide explains that related texts tend to be close in the embedding space. This supports finding passages that use different words to express the same topic or requirement. An illustrative knowledge-base query might ask how to recover access, while the relevant procedure uses a different phrase for the same operation. The application can compare the query's vector with stored passage vectors instead of relying only on exact word matching.
The maker's examples use unit-normalized embeddings. For such vectors, cosine similarity, dot product and Euclidean distance give consistent similarity rankings, even though their numerical values are not identical. Choose one suitable comparison method and apply it consistently across the query and document representations.
Similarity is not proof that two statements are interchangeable. Retain the source text and inspect the condition, date or identifier that matters to the question. Two passages can concern the same subject while describing different procedures or exceptions. The embedding provides a retrieval signal; interpreting the selected passage is a separate part of the task.
Retrieval and source-based answers
A practical document workflow prepares vectors for knowledge-base passages, stores their identifiers and original text, and then compares a new question with that collection. Return the relevant passages with their source locations. The application can show those passages directly or supply them to a text-generation model that writes an answer grounded in the retrieved material.
For a useful search result, retain the document title, section, passage and the identifier used to find it. A support assistant can then retrieve the procedure relevant to a user's question and pass that procedure to the answer-writing stage. Mistral Embed performs the numerical representation step; it does not itself draft the final explanation or execute an external action.
Other documented uses
The maker also lists paraphrase detection, classification, clustering and batched processing among embedding workflows. A comparison task can retrieve passages with related meaning for review. A classification workflow can compare an input with labeled examples or use the vectors as features for a separate classifier. A clustering workflow can group similar items for a later human or application-level description.
Define the purpose of the representation before preparing the data. A passage used for semantic search should preserve the information someone will need after retrieval; a classification example should include the wording that distinguishes the class. Keep the preparation consistent so that a query and a stored item represent comparable material.
Edition and output boundaries
Use the same model edition and compatible representation for the documents and their queries. A vector with the same number of coordinates from a different model is not automatically in the same semantic space. Replacing the model or representation requires the stored items and queries to remain aligned.
This entry describes text input and numerical vector output, with the maker's 8k context and Mistral Embed's documented 1,024-dimensional representation. The configurable dimensions and quantized formats of Codestral Embed belong to that separately named code model and should not be assumed here. Images, audio, generated prose and tool execution are separate capabilities outside this text-embedding result.
POST /v1/embeddingstext, token_idsembedding| Field | Type | Requirement | Values & limits | Default |
|---|---|---|---|---|
Fieldmodel | Typestring | RequirementRequired | Values & limits mistralai/mistral-embed-2312 | Default— |
Fieldinput | Typestring|string[]|integer[]|integer[][] | RequirementRequired | Values & limits Minimum items: 1 Conditions: Nonempty homogeneous batch. Token IDs are nonnegative integers. Input budget: Maximum: 8,192 · 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 | Contract8192 |
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 $0.0002 / 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": "mistralai/mistral-embed-2312", "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='mistralai/mistral-embed-2312',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 mistralai/mistral-embed-2312
Use the method paths shown above.