Alex Finn compares Opus 5 and Fable 5 on a roller coaster simulation, an Apple website reconstruction, a document scavenger hunt, a debugging wall and a bridge simulator. Opus 5 wins most of the testsA benchmark is a defined test or collection of tasks used to compare a system's performance under specified conditions. and frequently costs less for the completed taskCost per completed task measures the total AI, tool, infrastructure, retry, and repair expense for each verified useful outcome..
The model's weaknesses are less about raw capability than workflow fit. Alex Finn finds its personality verbose and unfocused, and argues that the surrounding coding harnessAn AI agent harness is the software framework that packages a model with tools, instructions, context management, execution controls, and user interaction. does not always help it use its intelligence efficiently.
The recommended approach is model routingAI model routing sends each request to a model chosen for that request's complexity, cost, speed, privacy, or other requirements. rather than a single permanent default. Opus 5 is reserved for difficult problems, Fable 5 can help with planning, and a faster general model remains useful for routine daily work.
Watch on YouTube




