Create Embeddings (Cohere)

POST/cohere/v2/embed

Generate embeddings for multimodal inputs (text, image, mixed) or when you need configurable embedding dimensions.

This endpoint is compatible with the Cohere Embedding API. API syntax follows Cohere's official documentation.

info

This endpoint is only compatible with qwen3-vl-embedding-8b. For the bge-m3 model, use the OpenAI-compatible endpoint.

---

Request

Parameter	Type	Required	Description
inputs	array		Array of input objects containing content blocks (text, image, or mixed)
model	string	✅	Must be qwen3-vl-embedding-8b
input_type	string	✅	search_document, search_query or image
output_dimension	integer		Output embedding dimension
embedding_types	array		e.g. ["float"]

warning

Following values classification and clustering are not supported as input_type.

Supported dimensions

output_dimension possible values are: 128, 256, 384, 512, 768, 1024, 1536, 2048, 2560, 3072, 3584 and 4096.

---

Text embedding

Request Example — Text embedding

curl -X POST https://api.ig1.ai/cohere/v2/embed \
  -H "Authorization: Bearer $IG1AI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "inputs": [
      {
        "content": [
          { "type": "text", "text": "The food was delicious" },
          { "type": "text", "text": "the waiter was happy" }
        ]
      }
    ],
    "model": "qwen3-vl-embedding-8b",
    "input_type": "search_document",
    "output_dimension": 2048,
    "embedding_types": ["float"]
  }'

Response example — Text embedding

{
  "id": "2988",
  "embeddings": {
    "float": [
      [
        0.021447632,
        -0.0013963302,
        "...",
        -0.004244844,
        0.030830972
      ]
    ]
  },
  "texts": [
    "The food was delicious"
  ],
  "meta": {
    "api_version": {
      "version": "2",
      "is_experimental": true
    },
    "billed_units": {
      "input_tokens": 20
    }
  },
  "model": "qwen3-vl-embedding-8b",
  "usage": {
    "billed_units": {
      "input_tokens": 20
    }
  }
}

Response fields

Field	Type	Description
id	string	Request identifier
embeddings	object	Embedding vectors grouped by type (e.g. float)
texts	array	Input texts that were embedded
meta.billed_units.input_tokens	integer	Tokens billed for this request
model	string	Model used

---

Image Embedding

Request Example — Image embedding

curl -X POST https://api.ig1.ai/cohere/v2/embed \
  -H "Authorization: Bearer $IG1AI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen3-vl-embedding-8b",
    "inputs": [
      {
        "content": [
          {
            "type": "image_url",
            "image_url": {
              "url": "data:image/png;base64,<BASE64_ENCODED_IMAGE_DATA>"
            }
          }
        ]
      }
    ],
    "input_type": "image",
    "output_dimension": 2048,
    "embedding_types": ["float"]
  }'

tip

For image inputs, use "input_type": "image" and wrap the base64-encoded image in an image_url content block. The response includes images counter under billed_units.

Response Example — Image embedding

{
  "id": "3181",
  "embeddings": {
    "float": [
      [
        0.03056407,
        -0.024977088,
        0.035329435,
        "...",
        -0.031385683,
        -0.023662506
      ]
    ]
  },
  "texts": [],
  "images": [
    {
      "width": 1280,
      "height": 720,
      "format": "png",
      "bit_depth": 32
    }
  ],
  "meta": {
    "api_version": {
      "version": "2",
      "is_experimental": true
    },
    "billed_units": {
      "input_tokens": 898,
      "images": 1
    }
  },
  "model": "qwen3-vl-embedding-8b",
  "usage": {
    "billed_units": {
      "images": 1,
      "input_tokens": 898
    }
  }
}

Response fields

Field	Type	Description
id	string	Request identifier
embeddings	object	Embedding vectors grouped by type (e.g. float)
meta.billed_units.input_tokens	integer	Tokens billed for this request
meta.billed_units.images	integer	Number of Images that were embedded
model	string	Model used