Model Availability Is Plan-Dependent, Not Fixed (CCAO-F)

What is stable and what is not

Model selection has two layers that the CCAO-F exam is careful to separate, and it tests the separation at the evaluate level. One layer is the specific product surface: which models appear in your picker, which one is the default, and whether the app switches models automatically. The other is the underlying decision logic: fast for structured volume, balanced for general work, high-capability for complex stakes. The first layer moves; the second holds.

The whole skill is not confusing the two. The specific lineup depends on your plan and changes over time, so it is something to verify in the product at the moment you need it, not a fact to commit to memory. The three-tier decision logic, by contrast, is durable enough to build your judgement on. Knowing which layer to trust as permanent is what this knowledge point evaluates.

The moving parts: picker, default, switching

Three things vary and should be checked rather than assumed. The first is the model-picker lineup -- the set of models actually offered to you, which differs across plans. The second is the default model -- the one selected before you choose, which is a product decision, not a statement that it is the right tier for your task. The third is automatic model-switching behaviour, where the app may move between models under certain conditions; whether and how it does so is plan- and feature-dependent.

Because all three vary and evolve, treating any of them as a fixed, memorised fact is fragile. A lineup you learned last quarter may have changed; a default that was one tier may now be another; switching behaviour may differ on your plan from what a colleague describes. The reliable move is to verify these in-app at the point of use. That is a general property of the platform, in the same way that plan-dependent context and usage behaviour are things to confirm rather than assume.

Why the logic survives the churn

The reason the three-tier logic stays stable while the lineup churns is that the logic describes task requirements, not product SKUs. Tasks will always range from fast-and-structured to deep-and-ambiguous, and matching a task to a fast, balanced, or high-capability tier is a statement about the task, not about which model names happen to be shipping. As long as there is a spectrum of capability-versus-speed, the logic applies, whatever the tiers are called.

This is why even a new tier appearing above or alongside the three-tier frame does not invalidate the selection logic. A fourth tier simply extends the spectrum; the reasoning -- match capability to the task's structure, complexity, and stakes -- is unchanged. The exam deliberately pins the Haiku/Sonnet/Opus frame as the thing to learn, precisely because it is the stable abstraction. Specific model names and their exact characteristics are verify-at-use details layered on top of that durable frame.

What the CCAO-F exam trips candidates on

Two errors are tested. The first is memorising a specific model list as a permanent fact rather than treating tier characteristics as the stable takeaway. A question may hinge on an outdated lineup or a specific model name; the credited reading treats the specific list as changeable and the fast/balanced/high-capability characteristics as the durable point.

The second is assuming the tier a plan defaults to is automatically the correct tier for every task run on that plan. The default is a product setting, not task advice. A question may present someone accepting the default for a task whose profile clearly points elsewhere; the credited answer applies the decision logic and overrides the default when the task calls for it.

Common misreadings to avoid

How this shows up on the exam

Questions may bait with specific model names, outdated lineups, or a scenario where someone leans on the default. The reliable reading treats the specific offerings as verify-at-use product details and the fast/balanced/high-capability logic as the durable knowledge. Apply the logic to the task and do not let the default or a memorised list substitute for it.

This knowledge point builds on the unified decision logic and reinforces the tier spectrum as the stable abstraction. It also connects to cost as a usage-budget consideration, another area where plan specifics vary but the underlying reasoning does not.