NVIDIA NIM Vision Language Models API

openapi: 3.1.0
info:
  title: NVIDIA NIM Vision Language Models API
  description: >
    Vision-language model inference through the standard /v1/chat/completions
    surface. Image inputs are passed as `image_url` content parts inside the
    messages array — either HTTPS URLs or base64 data URIs. Supports NVIDIA
    NeVA, microsoft/kosmos-2, phi-3-vision, llama-3.2-90b-vision-instruct, and
    other VLMs in the NIM catalog.
  version: '2026-05-25'
  contact:
    name: NVIDIA Developer Support
    url: https://forums.developer.nvidia.com/c/ai-data-science/nemo-llm-service/
  license:
    name: NVIDIA AI Enterprise License
    url: https://www.nvidia.com/en-us/data-center/products/ai-enterprise/
servers:
  - url: https://integrate.api.nvidia.com
    description: NVIDIA-hosted NIM endpoint
  - url: http://localhost:8000
    description: Self-hosted NIM container default
security:
  - BearerAuth: []
tags:
  - name: Vision
    description: Multimodal vision-language operations
paths:
  /v1/chat/completions:
    post:
      summary: Create A Multimodal Chat Completion
      description: >
        Generate a chat completion against a vision-language model. Image content
        is supplied via `image_url` parts inside the user message; the response
        mirrors the standard chat completions schema.
      operationId: createVisionChatCompletion
      tags:
        - Vision
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/VisionChatRequest'
      responses:
        '200':
          description: Multimodal chat completion response.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/VisionChatResponse'
        '400':
          description: Invalid request (e.g. image too large or unsupported format).
        '401':
          description: Missing or invalid API key.
        '413':
          description: Payload too large.
components:
  securitySchemes:
    BearerAuth:
      type: http
      scheme: bearer
      bearerFormat: nvapi-...
  schemas:
    VisionChatRequest:
      type: object
      required: [model, messages]
      properties:
        model:
          type: string
          description: e.g. `meta/llama-3.2-90b-vision-instruct`, `microsoft/phi-3.5-vision-instruct`, `nvidia/neva-22b`.
        messages:
          type: array
          items:
            type: object
            properties:
              role:
                type: string
                enum: [system, user, assistant]
              content:
                type: array
                items:
                  type: object
                  properties:
                    type:
                      type: string
                      enum: [text, image_url]
                    text:
                      type: string
                    image_url:
                      type: object
                      properties:
                        url:
                          type: string
                          description: HTTPS URL or `data:image/jpeg;base64,...`.
        max_tokens:
          type: integer
        temperature:
          type: number
        stream:
          type: boolean
    VisionChatResponse:
      type: object
      properties:
        id:
          type: string
        object:
          type: string
          example: chat.completion
        choices:
          type: array
          items:
            type: object
            properties:
              message:
                type: object
                properties:
                  role:
                    type: string
                  content:
                    type: string
              finish_reason:
                type: string
        usage:
          type: object
          properties:
            prompt_tokens:
              type: integer
            completion_tokens:
              type: integer
            total_tokens:
              type: integer