Fable Burns $200 in 2 Hours, User Switches to GPT 5.5

yacineMTB · x · 2026-07-05

yacineMTB complained that Fable (Anthropic) burned $200 in just two hours, prompting a switch to GPT 5.5. To his surprise, GPT 5.5 is "actually not that bad" and can handle the same tasks at a much better cost-performance ratio. This offers a real-world data point comparing the cost and usability of these two frontier models.

Related event: Anthropic's Fable Model Reported to Incur Extremely High Usage Costs(3 posts)→

Original post →

More from Models

Models channel →