A developer asks the most powerful frontier model to rewrite a button label.
That is the AI equivalent of lighting a cigar with a flamethrower. It works. The cigar is lit. The table is also on fire, the curtains are nervous, and nobody can remember why this needed military equipment.
The same thing happens with document classification, small copy edits, simple extraction, and the kind of code change where the answer is visible before the model starts thinking. A huge model will do these tasks. So will a smaller one. The difference is that the huge model brings more latency, more cost, more rate limit pressure, and sometimes more policy machinery than the task can use.
The sharp rule is simple: use the smallest model that completes the task reliably.
Not the smallest model in theory. Not the cheapest model on a pricing page. The smallest model that passes the actual test in front of it.
The Fable lesson is not about fear
Fable 5 and Mythos 5 made this visible because they were unusually powerful and unusually gated. Anthropic launched both on June 9, 2026. On June 12, the company said a US government export control directive forced it to suspend access to both models for all customers, while other Anthropic models stayed available.
That is a dramatic trigger, but the daily lesson is quieter.
Fable was not just another model in a neat ladder. Anthropic described it as a Mythos class model, above Opus, built for long and complex work. It said Fable could work more autonomously, with fewer with fewer interruptions, and that Mythos 5 was the same underlying model with certain safeguards lifted for vetted users.
That matters. It means the top tier is no longer just a smarter autocomplete box. It is capability plus access rules, retention rules, safety gates, and possible interruption from outside the vendor. The dropdown is starting to look less like a tool menu and more like airport security.
Using that tier to clean up marketing copy is not brave. It is untidy.
Most work has a ceiling
Anthropic’s own Claude model guide says Haiku is for quick answers, extraction, categorization, and simple summaries. Sonnet is positioned as the daily driver for coding, writing, analysis, research, and multi step workflows. Opus is reserved for deep research and complex reasoning, especially after Sonnet struggles. Fable is described as the heaviest option for the largest and most important long running tasks.
That is not a cost saving slogan. It is a control system.
Keep an escalation ladder
A sensible ladder starts low.
For bounded, repeatable, privacy sensitive work, start with a local or small model if it passes the evaluation. This is not because local AI is morally superior. It is because some jobs are small, private, repetitive, and better kept close to the machine.
For routine cloud work, use Haiku or Sonnet. Much routine cloud work will stop there. Fast enough, capable enough, less drama.
Escalate when the task earns it. A vague product refactor across a large codebase may earn it. A long running agent that must plan, edit, run tests, recover from failures, and keep the whole goal in view earns it. Security work in a messy repository may earn it. A migration that would take a team weeks may earn it.
Anthropic’s Fable launch examples point exactly there: long autonomous software engineering, large codebase changes, deeper document reasoning, hard scientific work, and specialized cyber or biology use. Those are not button labels. Those are the jobs where more capability can change the outcome, not just decorate the answer.
The mistake cuts both ways.
Using the frontier tier for everything is wasteful and brittle. Throwing it away because most work does not need it is just as wrong. The top model should remain available as an escalation tier, with clear rules for when it gets called.
You do not light a cigar with a flamethrower. You also do not throw away the flamethrower because you mostly smoke cigars.