Pat Simmons compares Gemini 4 Carbon with competing models on a website clone involving 3D product renders, a motion-design film and a browser flight game. He scores the outputs by his own visual and functional assessmentEvaluation is the systematic assessment of a system's behaviour or results against explicit criteria..
He reviews model benchmarks and introductory pricing, then compares the time, tokens and estimated API costs of the buildsCost per completed task measures the total AI, tool, infrastructure, retry, and repair expense for each verified useful outcome.. Carbon ranks last in his first two tests and third in the flight game, while using less money in those examples.
Simmons regards the model as an improvement over its predecessor but notes that his large creative builds are not a complete measure of everyday model performance. An additional multi-agent Eiffel Tower reconstruction appears at the end.
Watch on YouTube
