Claude certification domain weights: what each exam actually rewards

Claude certification domain weights are set out in the Exam Content Outline of each official guide, and they are not close to even. The largest single domain is Applications and Integration on the Developer Foundations exam at 33.1% — about a third of that paper. The smallest sits on the same exam: Eval, Testing, and Debugging at 2.6%, which works out to a single scored item. Studying every domain equally is wrong on all four tracks, and it is worst on the developer one.

Where claude certification domain weights come from

Each of the four exams has an official guide, and each guide contains an Exam Content Outline that assigns every domain an approximate share of the scored items. That outline is the only authoritative source for weighting. It is not inferred from question banks, community reports or the length of a syllabus section, all of which will mislead you.

Three of the four publish whole percentages. The Developer Foundations guide is the exception and publishes to one decimal place, which is why its figures below keep their decimals. Rounding them for visual tidiness would collapse 2.6% and 3.1% into the same-looking number and lose the most useful thing that guide tells you.

All four blueprints, with item counts

The item column converts each weight into an approximate number of questions on that exam's paper. It is the more useful column, because “12%” is abstract and “seven questions” is a study decision. Weights are approximate and exams include unscored items, so read the counts as close rather than exact.

The widest gap is on the developer exam

The interesting number is not any single weight but the distance between a track's heaviest and lightest domain. On Architect Foundations that distance is small: five domains between 27% and 15%, so no domain can be safely skipped and none dominates. On Developer Foundations it is enormous.

The other end of the scale is worth naming too: Agentic Architecture and Orchestration is the heaviest single domain on any of the four exams, which is why coordinator and subagent patterns repay study time better than anything else on the architect track.

On Developer Foundations, Applications and Integration is worth about seventeen questions and Eval, Testing and Debugging is worth one. Claude Code is worth two. A candidate who spends a week on Claude Code configuration because it is the part they use daily has spent a week on 3.1% of the exam — the same material is worth 20% on the architect track. That is the single most common way to fail a paper you knew the material for.

A weight is not a difficulty rating

A heavy domain is not necessarily a hard one, and a light domain is not necessarily easy. Weight tells you how many marks are available, not how likely you are to win them. Those are different questions and they combine: the domain that deserves your next hour is the one where a large share of the marks meets a measured weakness.

This is why a blueprint alone cannot produce a study plan. It supplies one axis. The other has to come from evidence about your own answers — which is what a timed mock test produces and a confident reading of a chapter does not.

Turning a weight into a study decision

Put the two axes together and the ranking falls out. Take your measured accuracy per domain from a full timed attempt, set it against that domain's weight, and work the quadrants in order.

The order matters more than the arithmetic. A domain worth 27% where you are scoring 60% has far more recoverable marks than a domain worth 7% where you are scoring 50%, even though the second gap looks worse. Fixing the first is worth roughly four times as much. For how those raw percentages relate to the reported result, see the scaled-score guide.

What the weights do not tell you

Three limits are worth stating plainly. First, a domain weight is not a per-domain pass mark: all four exams produce one scaled score against one standard, so a weak domain can be carried by strong ones. Second, weights are approximate proportions of scored items, and papers include unscored items, so the derived counts are close rather than promised.

Third, and most often missed, a weight says nothing about what inside a domain is examined. Each guide also publishes an out-of-scope list, and material can sit squarely inside a domain's subject while being excluded from the exam — retrieval implementation details and vision capabilities are both named exclusions on Architect Foundations. For what falls outside the lines, see what the exams deliberately leave out, and for the subject matter itself, what the Claude exam actually tests.

If you have not settled on a track yet, the certification comparison guide is the place to start, and the practice exam guide covers what is and is not available to rehearse with. Official outlines for each track are published on the Anthropic Partner Academy certification pages.

Key takeaways

• The weights are published, so use them. Every official guide contains an Exam Content Outline giving each domain an approximate share of scored items.
• The four exams are weighted very differently. Ranges run from 1.8 to 1 on Architect Foundations up to roughly 13 to 1 on Developer Foundations.
• Convert weights into questions before deciding anything. “2.6%” is abstract; “one item out of 53” changes how long you spend on it.
• Weight is only one axis. Rank domains by weight combined with measured accuracy from a timed attempt, not by weight alone and not by what interests you.
• A weight is not a pass mark. All four exams are scored as a single scaled score, so no individual domain has a threshold you must clear.