Charon

Benchmark

Past cases

Same model. Same prompt. Different interface. This case compares native Canva MCP with Charon on a precise infographic recreation.

Precise infographic recreation

Same model (5.6, high). Same prompt. The only difference is the API control interface.

Prompt
Recreate this infographic in Canva while making it completely editable.
Original Canva infographic used as the reference for the recreation task

Original

Infographic produced with Charon using Canva’s programmable Design Editing API

With Charon (Every single element is editable)

8 minutes

Infographic produced with native Canva MCP after four tries

With native Canva MCP

After 4 tries

14 minutes

Canva MCP File

Opened for editing. The result is almost entirely a single image, so the design is not meaningfully editable.

Canva MCP file opened for editing, showing a result that is almost entirely a single flattened image rather than editable objects

Opened in Canva

Earlier ChatGPT attempts

Four inaccurate outputs before the run was even in the right ballpark.

Four inaccurate ChatGPT Canva attempts before the result was in the right ballpark

Four attempts

What we are measuring

  • Complete task success
  • Visual and informational accuracy
  • Execution time
  • Model round trips
  • Human corrections required