<aside> 🛠

Evaluate Text Outputs from Multiple Models Side-by-Side

Load up to 4 text outputs and compare them in a synchronized side-by-side view. Perfect for evaluating responses from different LLMs / multimodal models, A/B-testing prompt variants, or checking how the same input is interpreted by different model families. Read all answers at once, spot differences fast, and pick the best output without copy-pasting between tabs.

</aside>

What it does

Text Compare displays up to 4 text outputs in a side-by-side view. Feed it the outputs from multiple models (e.g., Gemma 4 31B and Qwen3.6-35B-MoE responding to the same prompt) and the panels populate so you can read them in parallel. Compare answer quality, factual accuracy, tone, completeness, or formatting instantly. Essential for model evaluation, prompt tuning, and routing decisions in any text-generation workflow

Problem it solves

• Compare LLM outputs – Read responses from different language / multimodal models against the same prompt
• Prompt A/B testing – Run the same model with different prompts and compare results in one view
• Model selection – Pick the best model for a task by seeing all candidate answers at once
• Quality assessment – Spot hallucinations, factual differences, missing details, or formatting issues by direct comparison
• Tone & style comparison – Evaluate which model's writing style fits your use case
• Multi-model evaluation pipelines – Feed N model outputs into one node instead of running them one by one
• Faster iteration – Eliminate copy-pasting between separate output windows
• Visual decision-making – Make informed choices about routing, deployment, or prompt design

Input/Output

<aside>

• Input: Up to 4 text outputs
  • Source: Outputs from any text-producing node (Language Models, Multimodal Language Models, OCR results, text files, etc.)
  • Format: Plain text (each panel accepts one text stream)
  • Connection: Connect each model/text source to one of the 4 input ports on the left side of the node
  • Empty ports: Unconnected panels show a placeholder file icon and stay blank in the comparison view </aside>

<aside>

• Output: Synchronized multi-text view
  • Format: Side-by-side panels (up to 4)
  • Behavior: Each panel renders the text from its corresponding input port
  • Interaction: Read, scroll, and compare panels in parallel </aside>

How to Use

<aside>

1. Build your flow → Set up your text sources. Common pattern: a single Text Input feeding the same prompt to multiple LLM / multimodal nodes (e.g., Gemma 4 31B and Qwen3.6-35B-MoE), each producing a separate text output.

2. Add Text Compare → Drop a Text Compare node onto the canvas and connect each model's text output to one of the 4 input ports on the left

3. Run Flow → Each model runs and its response populates the corresponding panel in Text Compare

   

4. Read side-by-side → Open the Text Compare node to view all responses in parallel and pick the best one (or take notes on differences for further prompt/model tuning) </aside>

Features

• Multi-Image Display:
  • Load up to 4 text outputs – Compare 2, 3, or 4 model responses at once
  • Panel layout – Each connected input gets its own panel in the comparison view
  • Empty placeholders – Disconnected panels show a file-icon placeholder so the layout stays predictable