Pat Simmons gives GPT-5.6 Sol and Fable 5 the same prompts for three builds: a close recreation of a design tool, a deliberately unusual promotional concept and an interactive New York City learning experience. Each model initially receives a one-shot prompt with the same constraints and deployment requirement.
Sol is much slower on the first task but produces a substantially more complete recreation with deeper interactions. It also wins the creative-concept test by building a richer interactive experience, while Fable's result is clearer but more static. Both models need another round of direction before the learning platform becomes genuinely immersive.
Pat Simmons concludes that Sol followed directions more thoroughly and delivered better design and creative work in these tests. The reported cost figures are less conclusive because subscription accounting and local usage tools disagree, so the comparison is strongest as a workflow and output-quality test rather than a precise price benchmark.
Watch on YouTube
