Fable 5.1 cuts cache costs 75% but thinking-heavy tasks cost 20% more

AI model Fable 5.1 launched with a significant reduction in cache read pricing, dropping from $1.00 to $0.25 per million tokens, while input and output prices remained unchanged at $10 and $50 per million tokens respectively. The developer behind the analysis found that the cost impact depends heavily on workload type: agentic sessions with repeated context re-reads can see bills drop by up to 45%, while high-effort single-shot tasks cost significantly more. Independent benchmarking firm Artificial Analysis recorded a 20% cost increase per task over Fable 5, measuring at max effort settings where the cache savings are less impactful. A separate test by researcher Simon Willison showed that running the same prompt at maximum versus low effort could cost 33 times more, highlighting how the model's effort dial drives costs far more than token pricing. Subscription users on the Pro plan also faced surprise charges, as Fable 5.1 falls outside standard usage limits and runs on pay-as-you-go credits with no introductory credit offered at launch.