Introduction
In the world of AI, each technological advancement brings us closer to a future where machines can rival humans in complex creative tasks. This time, we challenged some of the most advanced AI models to see if they could "draw" the Mona Lisa, one of the most iconic artworks of all time. The models in question are GPT-5.6 Sol, Claude Fable 5, Grok 4.5, and Gemini 3.6 Flash.
Experiment Context
The idea was simple yet ambitious: give each model a blank canvas and a set of colored-pencil tools, then let them reproduce the Mona Lisa or draw from a simple text prompt. The goal was twofold: assess the models' ability to handle an open-ended task and observe the costs and time required to achieve acceptable results.
Tools and Methodology
Each model had the same tools:
- Plan: a no-op scratchpad for thinking and planning between steps.
- View_target: to look at the target image again.
- View_canvas: to see their own work and decide what to fix.
The models could adjust color, tip width, and pressure, lay down strokes, blend colors, and erase.
Results and Analysis
GPT-5.6 Sol
GPT-5.6 demonstrated impressive capability in handling the fine details of the Mona Lisa, but at a high cost in processing and time. The model produced visually convincing results but required multiple iterations to reach a satisfactory level.
Claude Fable 5
Despite high expectations, Claude Fable 5 took much longer and more financial resources than its competitors, while producing inferior results. This highlights a crucial point: a model's performance can vary significantly across tasks.
Grok 4.5
Grok struggled with this task, showing the limitations of open-weight models in complex tasks. Its results were often unfinished or lacking detail, highlighting the capability gap between cutting-edge models and more accessible ones.
Gemini 3.6 Flash
Gemini managed to balance speed and accuracy, producing satisfactory results at a reasonable cost. This shows that some models can offer a good compromise between cost and quality.
Implications and Reflections
This experiment highlights several key points. First, current AI models can indeed tackle creative tasks, but with varying levels of success. Second, cost and time remain critical factors to consider. Finally, this experiment underscores the importance of choosing the right model based on the specific task and project constraints.
Conclusion
The exercise of "drawing" the Mona Lisa with AI models not only reveals the current capabilities of these models but also the challenges to be overcome to achieve a level of performance truly comparable to human artists. So, where do we go from here?
Let's discuss your project in 15 minutes.