The Haiku / Sonnet / Opus Tier Spectrum for CCAO-F

One family, arranged as a spectrum

The Claude model family is not a grab-bag of unrelated options; it is a spectrum with a clear direction. The CCAO-F exam introduces this at the remember level because everything else in model selection depends on holding the ordering steady in your head. Three tiers -- Haiku, Sonnet, Opus -- run from fast-and-efficient at one end to thorough-and-capable at the other, with a balanced middle.

The reason to think of it as a spectrum rather than a menu is that the tiers trade one thing for another in a consistent direction. As you move up, you gain capability on complex work and give up speed; as you move down, you gain speed and give up some ceiling on nuanced work. That single directional trade is the mental model the whole domain builds on, which is why it unlocks the individual task profiles and the decision logic that combine them.

The three tiers and their positions

Haiku sits at the fast, efficient end. It is built for speed, which makes it a natural fit for structured work that runs at volume, where quick turnaround matters more than deep interpretation. Sonnet sits in the middle as the balanced, general-purpose tier -- the one that handles most everyday professional work well without leaning to either extreme. Opus sits at the thorough, high-capability end, built for depth of reasoning on genuinely complex or ambiguous work, accepting slower responses as the cost of that depth.

The positions matter because they are relative and ordered, not just three labels. Haiku is faster than Sonnet, which is faster than Opus; Opus is more capable on complex work than Sonnet, which has more headroom than Haiku. Knowing the order lets you reason about any task by asking where on the spectrum its demands fall, rather than memorising a fixed model-to-task table. The detailed fit of each tier is covered in the Haiku, Sonnet, and Opus profiles.

Why "higher" is not "better"

The spectrum's most important lesson is that up is not the same as better. Because Opus is the most capable tier, it is tempting to treat it as the safe, correct default for anything that matters. But the spectrum is a trade, not a ranking of quality: the capability Opus adds comes bundled with slower responses, and for a task that does not need that capability, the slower response is pure cost with no benefit.

This is why tier choice is about task fit, not tier prestige. A high-volume classification job does not become better on Opus; it becomes slower for no gain, while Haiku would clear it quickly. A nuanced, high-stakes judgment does not become adequate on Haiku just because Haiku is fast; it needs the depth further up the spectrum. The right tier is wherever the task's real demands fall, and the tiers differ in the kind of task each suits, not merely in price. The continuous nature of that trade is developed in the speed-versus-capability tradeoff.

What the CCAO-F exam trips candidates on

Two assumptions are tested. The first is assuming the tiers differ only in price and not in the kind of task each is suited for. The credited understanding is that the spectrum is about task fit -- speed versus depth -- and price is downstream of that, not the defining difference.

The second is assuming a higher tier is strictly better for every task regardless of what the task actually requires. The spectrum is a directional trade, so a higher tier is better only when the task needs the extra capability enough to justify the slower response. For structured, high-volume work, the higher tier is worse in practice because it is slower without improving the outcome. Both traps stem from treating the spectrum as a quality ranking rather than a trade.

Common misreadings to avoid

How this shows up on the exam

Questions ask you to order the tiers or to reject the idea that a higher tier is always the right pick. Anchor on the spectrum: Haiku fast, Sonnet balanced, Opus thorough, with a consistent speed-for-depth trade. The reliable answer treats tier choice as task fit, never as reaching for the most capable option by default.

This foundational knowledge point unlocks the three task profiles and the unified decision logic that combines them, and it grounds the speed-versus-capability tradeoff that turns the spectrum into a selection rule.