🎯 Task-SpecificVerified

Plate Parser - AI Menu Digitalizer Model

Menu Digitalizer

by Neurometric📄 Document Intelligence

Plate-Parser is a task-specific model that converts physical restaurant menu photos into structured data (Item, Price, Diet).

Plate Parser is designed for Document Intelligence workflows where speed, consistency, and control matter more than generic creativity. Instead of relying on a broad model to guess intent each time, this task-specific SLM is optimized for menu digitalizer and predictable output quality. That makes it easier to adopt in production pipelines where teams need reliable formatting, lower latency, and reduced hallucination risk.

Most teams integrate CRM records, internal docs, and workflow tools to automate repetitive language-heavy tasks with predictable outputs. This makes deployment practical for sales ops, customer success, legal, and operations teams that need answers grounded in their own data. Because the model is small and focused, teams can run it with efficient infrastructure while still meeting quality targets for the use case.

Compared with general-purpose models, focused SLMs like this one typically deliver better precision for narrow tasks, lower cost per request, and easier governance. You can start with the free tier, test the model against your business data, and scale usage through Neurometric APIs or ClawPack workflows when performance meets your benchmark.

Q

Model: Qwen/Qwen3-4B-Instruct-2507

This task-specific SLM uses Qwen/Qwen3-4B-Instruct-2507 with a specialized system prompt. The model weights are identical to the base Qwen model — the task specialization comes from the system prompt and has been validated through our testing pipeline. Need better accuracy? We can fine-tune the model on your data for your specific use case.

Quickstart

Copy-paste code to use Plate-Parser in your project

from openai import OpenAI

client = OpenAI(
  base_url="https://api.neurometric.ai/v1",
  api_key="YOUR_API_KEY"
)

resp = client.chat.completions.create(
  model="neurometric/plate-parser",
  messages=[
    {"role": "user",
     "content": "{\n  \"language\": \"en\",\n  \"image_paths\": [\n    \"uploads/IMG_1042.jpeg\"\n  ],\n  \"currency_hint\": \"GBP\",\n  \"diet_labels_supported\": [\n    \"V\",\n    \"VG\",\n    \"GF\",\n    \"NF\"\n  ]\n}"}
  ]
)
print(resp.choices[0].message.content)
OpenAI SDK compatibleTry in Playground →

Test Results

Example Input

Language: en Image Paths: - uploads/IMG_1042.jpeg Currency Hint: GBP Diet Labels Supported: - V - VG - GF - NF

Expected Output

Menu Items:

Diet: VG; Item: Avocado Toast; Price: 8.25

Diet: NF; Item: Smoked Salmon Bagel; Price: 9.75

Diet: V; Item: Buttermilk Pancakes; Price: 7.5

Actual Model Response

I'm unable to process or analyze images directly, including restaurant menu photos. The provided image_paths list contains a file path (uploads/IMG_1042.jpeg), but I don't have access to the actual image or the ability to extract text or structured data from it.

To convert a menu image into structured data (Item, Price, Diet), you would typically need to:

1.

Use an OCR (Optical Character Recognition) tool (like Google Vision, Tesseract, or AWS Textract) to extract text from the image.

2.

Parse the extracted text to identify:

Item names

Prices (with currency hint: GBP)

Diet labels (e.g., V = Vegetarian, VG = Vegan, GF = Gluten-Free, NF = No Fish)

3.

Validate and structure the output in a consistent format.

If you'd like, I can help you:

Design a schema for the output (e.g., list of dictionaries with Item, Price, Diet)

Write a Python script using libraries like pytesseract and Pillow (with OCR) to process such images

Provide examples of how the structured data might look

Let me know how you'd like to proceed!

Test this model

Sign in to run your own test cases against this model.

Capabilities

templatingtransformationspecialized-task

Frequently asked questions

Does Plate Parser work with CRM systems?

Yes. Plate Parser is built to work with CRM systems, internal docs, and CSV/JSON pipelines. Most teams connect existing APIs or scheduled exports and map records into prompt-ready JSON.

What data format does this model accept?

You can send plain text or structured JSON. For menu digitalizer, teams usually pass normalized records with field names, context notes, and output constraints.

Why use a task-specific SLM for menu digitalizer?

Task-specific SLMs are tuned for one business outcome, so they are often more consistent and cost-efficient than general models for repetitive production workflows.

Model Access

From $3/mo

Unlimited API calls

Subscribe

Or subscribe to multiple models from $7/mo

Try This Model

💬

Send a message to try the model

Use via API

Free

View API settings and get a free key. Sign up when you want more keys or billing access.

Need better performance?

We can fine-tune this model on your data for higher accuracy on your specific use case.

Model Info

ModelQwen/Qwen3-4B-Instruct-2507
Parameters4B
Last Updated2026-03-19