GMAT Focus replaced the legacy GMAT in late 2023. The shorter format, the dropped AWA, the renamed Data Insights section — most candidates have absorbed the basics by now. What they haven’t absorbed is that GMAC continues to retune the test on a quarterly cadence. The shape of the exam in early 2026 is not the shape it had in mid-2024, and the gap is starting to matter.

We track these revisions because our engine has to. When GMAC adjusts item parameters, official mock calibration, or section emphasis, every adaptive prep platform that calibrates against the published spec inherits the change — or fails to. This brief is the diff.

What actually shifted

Four observable changes stand out in the Focus 2026 specification. None of them are dramatic on their own. Taken together, they reshape what a high-band score actually rewards.

1. Data Insights weighting has crept upward

GMAC has gradually shifted the contribution of Data Insights toward the composite Total Score. In the early Focus rollout, candidates and consultants treated DI as the “new section” — important, but mentally bucketed alongside the dropped IR. That framing is no longer accurate. DI now carries genuine weight in the Total, and the published percentile distributions for DI have tightened meaningfully at the top end. A 90th-percentile DI score in 2024 is closer to an 84th-percentile DI score against the current curve.

Translation: the section that most candidates still underprepare is the section where marginal improvement now buys the most.

2. Multi-source reasoning is the load-bearing DI sub-skill

Within Data Insights, multi-source reasoning items are no longer the niche cousin of table analysis and graphic interpretation. The proportion of MSR items on retired official mocks has crept up, and the items themselves have gotten denser — three tabs of mixed prose, tables, and charts is now closer to the median MSR stimulus than the upper bound.

This rewards a specific habit: building a mental schema of the entire stimulus before touching the first sub-question. Candidates trained on legacy IR pacing skim, jump to the question, and then re-read. That loop is now expensive.

3. Pacing tolerance has narrowed at the top of the band

Official practice mocks have been re-calibrated such that the time penalty for a misallocated minute is steeper near the ceiling. This is partly a mechanical consequence of fewer items per section — every question carries more weight — and partly a deliberate scoring choice. The practical effect: a 705-band candidate who burns ninety seconds on a single quant item now slips further than the same candidate would have in 2024.

4. Verbal item style has tilted toward inference

Sentence correction is gone — that part of the transition is settled. What’s newer is the tilt within Critical Reasoning and Reading Comprehension toward inference and assumption-style items, away from straightforward main-idea and strengthen/weaken work. The shift is subtle on any single mock, but the trend across the last several official releases is consistent.

Why most prep is still calibrated for the old shape

The major prep publishers update their question banks on a roughly two-year cycle. Most flagship courses on the market in 2026 were authored against the 2023–2024 spec. That’s not negligence — it’s economics. Rewriting a several-thousand-item bank costs more than most providers can recoup in a single revision cycle.

The consequence for candidates is straightforward. Pacing instincts are wrong. DI is under-weighted in study plans. MSR is treated as a curiosity rather than the section’s center of gravity. And the “feel” of a high-band item — what a 715-level CR stem actually looks like in 2026 — drifts further from the practice material every quarter.

What we updated

Our adaptive engine re-weights item parameters whenever GMAC publishes a calibration shift. For Focus 2026 specifically, three things changed under the hood.

First, the mastery vector — the internal representation of where a candidate stands on every sub-skill — now carries finer-grained Data Insights dimensions. We split MSR into three sub-dimensions (schema-building, cross-tab synthesis, and quantitative extraction) because the old single-axis representation was losing signal.

Second, the pacing model now penalizes minute-level drift more aggressively in the upper bands. In our internal testing, candidates whose pacing profile matched the new official-mock distribution converged on their target band materially faster than candidates running on the old pacing prior.

Third, the difficulty estimates on roughly a third of our Verbal bank were re-fit to reflect the inference tilt. Items that were calibrated as “medium” in 2024 because they had clean strengthen-style structure are now reclassified — the test no longer rewards that pattern at the same rate.

If you’re sitting Focus 2026, do this

  1. Spend a real week on Data Insights. Not a token weekend. If your most recent mock had DI sub-600, DI is your highest-leverage section, full stop. Multi-source reasoning is where the time goes.
  2. Take a fresh diagnostic on current official material.Mocks older than nine months are calibrated to a test that no longer exists in the same shape. Your “baseline” from last summer is not your baseline today.
  3. Train pacing on the new clock, not the old one. Practice in section-timed mode. The 2024 instinct to bank ninety seconds on the easy items and spend them on the hard ones is now a worse trade than it used to be — the easy items got faster too.

Closing note

We’ll publish a brief like this whenever a real test-spec change lands. Not for every press release — most GMAC updates are administrative — but whenever the underlying shape of the exam moves enough to change how a serious candidate should prepare. If you’re sitting Focus in the next six months, the right mental model is this: the test is incrementally more DI-weighted, incrementally faster, and incrementally less tolerant of stale prep. Plan accordingly.