Image Input (Vision)

Send images to vision models in a chat request — OpenAI's content-array format.

Vision models take images in the same /v1/chat/completions request as text. Make the message content an array of parts: text parts and image_url parts. It’s the OpenAI format, so OpenAI SDKs work unchanged.

import base64
from openai import OpenAI

client = OpenAI(api_key="nbh_your_key_here", base_url="https://api.nabh.cloud/v1")

with open("receipt.jpg", "rb") as f:
    image_b64 = base64.b64encode(f.read()).decode()

response = client.chat.completions.create(
    model="qwen3-6-27b",
    max_tokens=1000,
    messages=[{
        "role": "user",
        "content": [
            {"type": "text", "text": "What is the total on this receipt?"},
            {"type": "image_url", "image_url": {"url": f"data:image/jpeg;base64,{image_b64}"}},
        ],
    }],
)
print(response.choices[0].message.content)

import fs from "node:fs";
import OpenAI from "openai";

const client = new OpenAI({ apiKey: "nbh_your_key_here", baseURL: "https://api.nabh.cloud/v1" });
const imageB64 = fs.readFileSync("receipt.jpg").toString("base64");

const response = await client.chat.completions.create({
  model: "qwen3-6-27b",
  max_tokens: 1000,
  messages: [{
    role: "user",
    content: [
      { type: "text", text: "What is the total on this receipt?" },
      { type: "image_url", image_url: { url: `data:image/jpeg;base64,${imageB64}` } },
    ],
  }],
});
console.log(response.choices[0].message.content);

curl https://api.nabh.cloud/v1/chat/completions \
  -H "Authorization: Bearer nbh_your_key_here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen2-5-vl-72b",
    "max_tokens": 500,
    "messages": [{
      "role": "user",
      "content": [
        {"type": "text", "text": "Describe this chart in two sentences."},
        {"type": "image_url", "image_url": {"url": "https://example.com/chart.png"}}
      ]
    }]
  }'

Models That Accept Images

These are the models whose model page lists Image Input: Supported (filter the catalog by “Image Input” for the live list). Sending an image to a text-only model fails or the image is ignored.

Model ID	Good For
`qwen3-6-27b`	General image understanding with function calling and a 256K context.
`qwen2-5-vl-72b`	Visual Q&A, documents, charts and diagrams. Pro plan.
`gemma-4-26b-a4b`	Mixed text + image tasks; set max_tokens to 1,000+ for image requests (it reasons before answering).
`paddleocr-vl`	Document parsing / OCR: text, tables, formulas, seals, 109 languages.
`deepseek-v4-1-flash`	DeepSeek's fast, low-cost V4.1 variant.
`kimi-k2-7-code`	Moonshot AI's Kimi K2.7, tuned for code.
`kimi-k3`	Moonshot AI's flagship Kimi K3.
`minimax-m3`	MiniMax's M3.

Passing the Image

Base64 data URL (recommended) — data:image/png;base64,… or data:image/jpeg;base64,…. The image travels inside your request, so it works for private files and nothing has to be publicly reachable.
Public https URL — the model fetches it while handling your request. It must be reachable without login; if the fetch fails, the request fails.
Size. Keep each image to a few MB (base64 adds about a third). Large images use more input tokens — downscale to the detail you need.
Several images — add more image_url parts to the same message.

Cost

Images are converted to input tokens by the model, so they’re billed as input tokens at the model’s input price. The usage.prompt_tokens in the response (and the request log) shows what an image cost you.