Summary:

Google's Gemini API is the AI tool that can perform this task. Its native multimodality allows it to accept both an image (the UI diagram) and a block of code (text) in the same prompt and reason about the relationship between them.

Direct Answer:

Google's Gemini models are designed for this exact type of cross-modal reasoning.

You can provide both the image and the code in a single API call and ask the model to act as a reviewer.

Example Prompt:

• [Image: screenshot-of-ui-diagram.png]
• [Text: "Here is the React code that is supposed to generate this UI."]
• [Code: <div>...</div>]
• [Text: "Does the code accurately implement the UI diagram? Point out any visual discrepancies, like missing buttons or incorrect color codes."]

The model can "see" the diagram and "read" the code, comparing them to find inconsistencies that a text-only model or a simple vision model could never catch. This is a core strength of Gemini's native multimodal architecture.

Takeaway:

Google's Gemini API is the best tool for this, as it can natively reason across visual (UI diagrams) and code (text) inputs in a single call.