Chris tested the newly released Fable 5 AI coding model and found it vastly superior to his usual models, solving complex and creative coding challenges in just one or two prompts where others failed. Despite its high cost and resource demands, he believes Fable 5’s exceptional capabilities justify paying full price for its return to efficiently tackle critical problems in his revenue-generating apps.
Last week, Anthropic released Fable 5, a highly powerful AI coding model that quickly gained significant hype before being pulled by the US government within 48 hours. The creator, Chris, tested Fable 5 extensively during that brief window, comparing it against his usual go-to models, Opus 4.8 and GPT 5.5, both set to their highest coding settings. Chris challenged Fable 5 with some of his toughest coding problems—tasks that previous models had failed to solve—and was impressed by its ability to handle complex scenarios in just one or two prompts, which convinced him that Fable 5 is the best coding model he has ever used.
One of the toughest problems Chris tested was porting a custom drag-and-drop system from iPhone to iPad for his app Ellie. The iPad version required handling multiple columns and various drag targets, creating over ten edge cases that previous models struggled to manage simultaneously. While Opus and GPT 5.5 failed to deliver a working solution after multiple attempts, Fable 5 solved the problem perfectly on the first try. This demonstrated Fable’s superior ability to think many steps ahead and manage complex, multi-faceted coding challenges.
Chris then tested Fable 5 on an even more complicated web drag-and-drop feature involving a fragile, legacy codebase with over 10,000 lines of code and more than 50 edge cases. The goal was to enable dragging tasks from a calendar back into a Kanban board, a feature he had wanted for years but never managed to implement successfully with other models. Remarkably, Fable 5 accomplished this in just two prompts, and after hours of rigorous testing and code review using Greptile, an AI-powered code review tool, no significant issues were found. This further solidified Chris’s confidence in Fable’s capabilities.
Chris also explored Fable’s creativity by asking it to make emojis rain down with physics effects when tapping a logo in his calorie tracking app Amy. Fable nailed the task in one prompt, including creative choices like using food-related emojis, which would have taken other models multiple attempts. He concluded that Fable excels particularly in complex, creative, and legacy code scenarios where anticipating many steps ahead is crucial, whereas other models perform adequately on simpler tasks requiring fewer steps of reasoning.
However, Chris highlighted a major catch: Fable 5 is extremely resource-intensive and expensive to use, especially at the ultra code setting, which quickly exhausts usage limits. This means it’s not practical to use Fable for every problem but rather reserved for the most challenging issues that other models cannot solve. Given its power and cost, Chris is willing to pay a premium—potentially thousands of dollars—to solve critical problems efficiently, especially since his apps generate revenue. Ultimately, he believes the hype around Fable 5 is justified and looks forward to its return, even if only available via costly API pricing.