Gemini 4 Pro Tests Yield Impressive Results
Google has been working on a new frontier-class model for its Gemini AI, and recent tests suggest it could be a game-changer. The company has been testing early checkpoints of Gemini 4 Pro under the codename 'argon', with results showing significant improvements over previous models.
The first test result was shared by developer Lentils, who posted an example of Gemini 4 Pro's output on X. The model took 2.4 minutes to produce a fully monographic website that looked clean and well-polished, unlike the robotic templates typically generated by AI tools.
Another developer, Bee, also shared his test results on X, featuring a monographic website created by Gemini 4 Pro in just 14 minutes. The final result was impressive, with a clean and polished design that looked like it had been created by a human.
The audio effects used to create the sound of sketching on canvas paper were also a neat touch that added to the interface's tactile feel. While the model took longer than expected, the results suggest that Google may have finally resolved the design taste issues that plagued earlier Gemini models.