openapi.yaml

openapi: 3.0.3
info:
  title: Portkey Models API
  description: |
    Comprehensive pricing and configuration data for 1000+ LLMs across 35+ providers.

    ## Pricing Unit

    > ⚠️ **Prices are in cents per token, not dollars.**

    | JSON | Per 1K | Per 1M |
    |------|--------|--------|
    | `0.003` | $0.03 | $30 |
    | `0.00025` | $0.0025 | $2.50 |
    | `1` | $10 | $10,000 |

    **Cost Calculation:**
    ```javascript
    const costDollars = (tokens * price) / 100;
    ```

    ## Resources

    - [UI Explorer](https://portkey.ai/models)
    - [GitHub Repository](https://github.com/Portkey-AI/models)
    - [Portkey AI Gateway](https://github.com/Portkey-AI/gateway)

  version: 1.0.0
  contact:
    name: Portkey Support
    url: https://portkey.ai
    email: support@portkey.ai
  license:
    name: MIT
    url: https://opensource.org/licenses/MIT

servers:
  - url: https://api.portkey.ai
    description: Production API

tags:
  - name: Pricing
    description: Model pricing configurations
  - name: Configuration
    description: Model capabilities and parameters

paths:
  /model-configs/pricing/{provider}/{model}:
    get:
      summary: Get Model Pricing
      description: |
        Returns pricing configuration for a specific model.

        **Note:** Prices are in USD cents per token.

        ## Supported Providers

        openai, anthropic, google, azure-openai, bedrock, mistral-ai, cohere, 
        together-ai, groq, deepseek, fireworks-ai, perplexity-ai, anyscale, 
        deepinfra, cerebras, x-ai, and 25+ more.

        ## Example Response Fields

        | Field | Description | Unit |
        |-------|-------------|------|
        | `request_token.price` | Input token cost | cents/token |
        | `response_token.price` | Output token cost | cents/token |
        | `cache_write_input_token.price` | Cache write cost | cents/token |
        | `cache_read_input_token.price` | Cache read cost | cents/token |
        | `additional_units.*` | Provider-specific features | cents/unit |

      operationId: getModelPricing
      tags:
        - Pricing
      parameters:
        - name: provider
          in: path
          required: true
          description: |
            Provider identifier. Use lowercase with hyphens.

            Examples: `openai`, `anthropic`, `google`, `azure-openai`, `bedrock`, `x-ai`
          schema:
            type: string
            example: openai
        - name: model
          in: path
          required: true
          description: |
            Model identifier. Use the exact model name as specified by the provider.

            Examples: `gpt-4`, `gpt-4o`, `claude-3-5-sonnet-20241022`, `gemini-2.0-flash-001`
          schema:
            type: string
            example: gpt-4
      responses:
        '200':
          description: Pricing configuration for the specified model
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/PricingConfig'
              examples:
                openai-gpt4:
                  summary: OpenAI GPT-4
                  value:
                    pay_as_you_go:
                      request_token:
                        price: 0.003
                      response_token:
                        price: 0.006
                    calculate:
                      request:
                        operation: sum
                        operands:
                          - operation: multiply
                            operands:
                              - value: input_tokens
                              - value: rates.request_token
                          - operation: multiply
                            operands:
                              - value: cache_write_tokens
                              - value: rates.cache_write_input_token
                          - operation: multiply
                            operands:
                              - value: cache_read_tokens
                              - value: rates.cache_read_input_token
                      response:
                        operation: multiply
                        operands:
                          - value: output_tokens
                          - value: rates.response_token
                    currency: USD
                openai-gpt4o-with-tools:
                  summary: OpenAI GPT-4o (with additional units)
                  value:
                    pay_as_you_go:
                      request_token:
                        price: 0.00025
                      response_token:
                        price: 0.001
                      cache_write_input_token:
                        price: 0
                      cache_read_input_token:
                        price: 0.000125
                      additional_units:
                        web_search:
                          price: 1
                        file_search:
                          price: 0.25
                    calculate:
                      request:
                        operation: sum
                        operands:
                          - operation: multiply
                            operands:
                              - value: input_tokens
                              - value: rates.request_token
                      response:
                        operation: multiply
                        operands:
                          - value: output_tokens
                          - value: rates.response_token
                    currency: USD
                anthropic-claude:
                  summary: Anthropic Claude 3.5 Sonnet
                  value:
                    pay_as_you_go:
                      request_token:
                        price: 0.0003
                      response_token:
                        price: 0.0015
                      cache_read_input_token:
                        price: 0.00003
                      cache_write_input_token:
                        price: 0.000375
                    currency: USD
                google-gemini:
                  summary: Google Gemini 2.5 Pro (with thinking tokens)
                  value:
                    pay_as_you_go:
                      request_token:
                        price: 0.000125
                      response_token:
                        price: 0.001
                      additional_units:
                        thinking_token:
                          price: 0.001
                        web_search:
                          price: 3.5
                        search:
                          price: 3.5
                    currency: USD
        '404':
          description: Model or provider not found
          content:
            application/json:
              schema:
                type: object
                properties:
                  error:
                    type: string
                    example: Model not found

  /model-configs/general/{provider}/{model}:
    get:
      summary: Get Model Configuration
      description: |
        Returns general configuration and capabilities for a specific model.

        ## Configuration Fields

        | Field | Description |
        |-------|-------------|
        | `params` | Parameter limits and defaults |
        | `type.primary` | Model type (chat, text, embedding, etc.) |
        | `type.supported` | Supported features (tools, image, etc.) |
        | `messages.options` | Supported message roles |

      operationId: getModelConfig
      tags:
        - Configuration
      parameters:
        - name: provider
          in: path
          required: true
          description: Provider identifier (e.g., openai, anthropic, google)
          schema:
            type: string
            example: openai
        - name: model
          in: path
          required: true
          description: Model identifier
          schema:
            type: string
            example: gpt-4o
      responses:
        '200':
          description: Model configuration
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/GeneralConfig'
              examples:
                openai-gpt4o:
                  summary: OpenAI GPT-4o
                  value:
                    params:
                      - key: max_tokens
                        maxValue: 16384
                      - key: response_format
                        defaultValue: null
                        options:
                          - value: null
                            name: Text
                          - value: json_object
                            name: JSON Object
                          - value: json_schema
                            name: JSON Schema
                        type: string
                    type:
                      primary: chat
                      supported:
                        - tools
                        - image
                anthropic-claude:
                  summary: Anthropic Claude 3 Opus
                  value:
                    params:
                      - key: max_tokens
                        maxValue: 4096
                    type:
                      primary: chat
                      supported:
                        - tools
                        - image
                        - cache_control
                    messages:
                      options:
                        - system
                        - user
                        - assistant
        '404':
          description: Model or provider not found

components:
  schemas:
    PricingConfig:
      type: object
      description: Complete pricing configuration for a model
      properties:
        pay_as_you_go:
          $ref: '#/components/schemas/PayAsYouGo'
        batch_config:
          $ref: '#/components/schemas/BatchConfig'
        calculate:
          $ref: '#/components/schemas/CalculateConfig'
        currency:
          type: string
          enum: [USD]
          description: Currency code (always USD)
        finetune_config:
          $ref: '#/components/schemas/FinetuneConfig'

    PayAsYouGo:
      type: object
      description: Token-based pricing (all prices in USD cents)
      properties:
        request_token:
          $ref: '#/components/schemas/TokenPrice'
        response_token:
          $ref: '#/components/schemas/TokenPrice'
        cache_write_input_token:
          $ref: '#/components/schemas/TokenPrice'
        cache_read_input_token:
          $ref: '#/components/schemas/TokenPrice'
        request_audio_token:
          $ref: '#/components/schemas/TokenPrice'
        response_audio_token:
          $ref: '#/components/schemas/TokenPrice'
        cache_read_audio_input_token:
          $ref: '#/components/schemas/TokenPrice'
        additional_units:
          type: object
          description: |
            Provider-specific additional pricing units.

            Common additional units:
            - `web_search`: Web search tool usage
            - `file_search`: File search tool usage
            - `thinking_token`: Chain-of-thought reasoning tokens (Google)
            - `image_token`: Image generation tokens
            - `video_duration_seconds_*`: Video generation (OpenAI Sora)
          additionalProperties:
            $ref: '#/components/schemas/TokenPrice'
        image:
          $ref: '#/components/schemas/ImagePricing'

    TokenPrice:
      type: object
      description: Price object (value is in USD cents)
      properties:
        price:
          type: number
          description: |
            Price in USD cents per token/unit.

            Example: `0.003` = 0.003 cents/token = $0.03 per 1K tokens

    BatchConfig:
      type: object
      description: |
        Batch API pricing configuration.
        Prices are specified as exact values matching the provider's published batch pricing.
      properties:
        request_token:
          $ref: '#/components/schemas/TokenPrice'
        response_token:
          $ref: '#/components/schemas/TokenPrice'
        cache_read_input_token:
          $ref: '#/components/schemas/TokenPrice'
      example:
        request_token:
          price: 0.000125
        response_token:
          price: 0.0005

    ImagePricing:
      type: object
      description: Image generation pricing by quality and size
      additionalProperties:
        type: object
        additionalProperties:
          $ref: '#/components/schemas/TokenPrice'
      example:
        standard:
          1024x1024:
            price: 4
          1024x1792:
            price: 8
        hd:
          1024x1024:
            price: 8
          1024x1792:
            price: 12

    CalculateConfig:
      type: object
      description: Cost calculation formulas
      properties:
        request:
          $ref: '#/components/schemas/CalculateOperation'
        response:
          $ref: '#/components/schemas/CalculateOperation'

    CalculateOperation:
      type: object
      description: Mathematical operation for cost calculation
      properties:
        operation:
          type: string
          enum: [sum, multiply]
          description: Operation type
        operands:
          type: array
          description: Operands for the operation
          items:
            oneOf:
              - $ref: '#/components/schemas/CalculateOperation'
              - $ref: '#/components/schemas/ValueReference'

    ValueReference:
      type: object
      properties:
        value:
          type: string
          description: |
            Reference to a value or rate.

            Examples:
            - `input_tokens`: Number of input tokens
            - `output_tokens`: Number of output tokens
            - `rates.request_token`: Request token rate
            - `rates.response_token`: Response token rate

    FinetuneConfig:
      type: object
      description: Fine-tuning pricing configuration
      properties:
        pay_per_token:
          $ref: '#/components/schemas/TokenPrice'
        pay_per_hour:
          $ref: '#/components/schemas/TokenPrice'

    GeneralConfig:
      type: object
      description: General model configuration and capabilities
      properties:
        params:
          type: array
          description: Model parameters with their constraints
          items:
            $ref: '#/components/schemas/ParamConfig'
        type:
          $ref: '#/components/schemas/ModelType'
        messages:
          $ref: '#/components/schemas/MessagesConfig'
        removeParams:
          type: array
          description: Parameters to remove for this model
          items:
            type: string
        disablePlayground:
          type: boolean
          description: If true, model is not available in Portkey playground
        isDefault:
          type: boolean
          description: If true, this is the default model for the provider

    ParamConfig:
      type: object
      description: Parameter configuration
      properties:
        key:
          type: string
          description: Parameter name
        defaultValue:
          description: Default value for the parameter
          oneOf:
            - type: number
            - type: boolean
            - type: string
            - type: 'null'
        minValue:
          type: number
          description: Minimum allowed value
        maxValue:
          type: number
          description: Maximum allowed value
        type:
          type: string
          description: Parameter type (e.g., boolean, string, array-of-strings)
        options:
          type: array
          description: Available options for enum-type parameters
          items:
            type: object
        skipValues:
          type: array
          description: Values to skip/ignore
          items: {}

    ModelType:
      type: object
      description: Model type classification
      properties:
        primary:
          type: string
          enum: [chat, text, embedding, image, audio, moderation]
          description: Primary model type
        supported:
          type: array
          description: Supported features
          items:
            type: string
            enum: [tools, image, cache_control]

    MessagesConfig:
      type: object
      description: Message configuration
      properties:
        options:
          type: array
          description: Supported message roles
          items:
            type: string
            enum: [system, user, assistant, developer]