Nate B. Jones tests Fable 5.1 on work that ends in inspectable artifacts rather than coding benchmarks alone. A real acquisition prompt produces a multi-sheet discounted-cash-flow workbook and a presentation with functioning formulas, explicit scenarios, sources and checks, despite incomplete public information about the target company.
Jones compares the same finance and writing tasks with competing models. He prefers Fable's presentation design and direct prose, while noting that another model makes audit trails and formula logic easier to inspect. The comparison emphasizes that a useful knowledge-work model must expose assumptions and create files that another person can review, not merely produce a confident answer.
The most striking test gives Fable one Seattle property address and asks it to create a film. Without Jones knowing Blender, the model researches the property, builds the house, terrain, lighting and camera path in code, checks rendered frames and delivers a 37-second sequence. Jones concludes that the low setting offers a practical balance of speed, cost and token use for routine professional work. Product and course promotions are omitted.
Watch on YouTube
