Anthropic's Fable 5 model costs twice as much as Opus 4.8, but the company's engineers have found a way to reduce costs by more than 50%. The secret lies in the 10-80-10 system, which redistributes the workload between expensive and cheap AI models.
The essence of the approach is simple: Fable 5 is used only where its high intelligence is truly needed. All other work is handled by cheaper models. This method, described as a "cheat code" for working with Fable 5, helps avoid wasting tokens on routine tasks.
How the 10-80-10 system works
The first 10% — planning. Fable 5 acts as the architect: it defines the structure, approach, success criteria, and constraints. Here, the model is not an executor but a strategist. If the plan is correct, the foundation for all work will be solid.
The main 80% — execution. Routine tasks, iterations, and minor edits are handed over to cheaper models: Opus 4.8 for standard tasks, Haiku for light ones, and for some cases, even GPT-5.5 will suffice. It is at this stage that the most tokens are consumed, so paying double the price for Fable is simply not cost-effective.
The final 10% — verification. Fable 5 returns to check the result against the original architecture: does it match the plan, and are there any gaps? Since this is a review of a completed result rather than generation from scratch, token consumption here is minimal.
Additional cost-saving tips
Analysts recommend starting not with the maximum but with a medium effort level: Fable on medium mode outperforms Opus on high mode. It is also advisable to delete old skills created for previous models — short instructions are cheaper. It is helpful to give models not only the task but also its reason: this reduces the number of edits and, consequently, token consumption. The /usage command can help track spending.
Costly mistakes are highlighted separately. The first is accidentally launching Fable: Claude Code and the app now default to this model, so the model selector should be checked every session. The second is the absence of a spending limit. Fable will soon switch to token-based billing, so it is worth setting a hard limit in the settings in advance.
My opinion: The 10-80-10 system is not just a way to save money but a strategic approach to managing AI resources. In an environment where token costs are rising, the ability to distribute the workload between models is becoming a critical skill for developers and analysts. The market is moving toward a point where AI usage efficiency will determine not only the budget but also the quality of the final product.