USMLE Step 3 · effective March 10, 2026

Step 3 has two format shifts, not one.

Most write-ups cover the March 10, 2026 block change: Day 1 went from 6 blocks of 38-39 items at 60 minutes to 12 blocks of 18-20 at 30, and Day 2 MCQ went from 6 blocks of 30 items at 45 minutes to 9 blocks of 20 at 30. Item content, total counts, and pass standard are unchanged.

What those write-ups skip is the other format shift that has always been on Step 3 and only Step 3: the same patient scenario can appear as a paragraph vignette or as a chart-tabular layout. Under 30-minute blocks, the time-to-parse delta between those two layouts is no longer negligible. Below is how the two shifts interact, and how to drill for both.

See the structural change →

Matthew Diakonov, Written with AI

Published May 20, 20269 min read

Direct answer · verified 2026-05-20

Vignette content and template are unchanged. The container around them is what shifted.

Effective March 10, 2026, Step 3 Day 1 runs as 12 blocks of 18-20 items at 30 minutes each (was 6 blocks of 38-39 at 60). Day 2 MCQ runs as 9 blocks of 20 items at 30 minutes each (was 6 blocks of 30 at 45). Total items, total day length, and pass standard are unchanged. The clinical vignette itself, its length, and its information density did not change. The other format shift, unique to Step 3, is internal to the items: the same patient scenario can render as a paragraph vignette or as a chart-tabular layout. With the block timer halved, the parse-time gap between those two layouts now matters. Source: USMLE test delivery software updates page.

The two things people mean by "Step 3 format shift"

When a coresident says they want to drill on the new Step 3 format, they could mean either of two things, and they probably have not separated them in their own head. It is worth separating them, because each one demands a different drill response.

The block-container shift (March 10, 2026). The exam delivery software changed and with it the block structure. Twice as many blocks on Day 1 with half the timer. Half again as many MCQ blocks on Day 2 with the timer compressed from 45 minutes to 30. This is the shift every commentary article is writing about.
The intra-item layout shift (always-on, only on Step 3). Step 3 is the only USMLE exam where the same patient scenario can be presented as a paragraph vignette or as a chart-tabular layout. This has been documented in the USMLE Step 3 sample items materials for years, but it gets cosmetic coverage at best. Under 30-minute blocks the parse-time delta between those layouts compounds into real seconds-per-block, which is why it suddenly matters more than it used to.

Below, both. First the block container, with the actual numbers. Then the chart-vs-paragraph layout split, with what to actually do about it during dedicated drill weeks.

Day 1: block container before and after

The left column is what every Day 1 examinee sat through up to March 9, 2026. The right column is what every Day 1 examinee sits through starting March 10. Total item count holds at 232, total day length holds at approximately 7 hours.

Feature	Through March 9, 2026	From March 10, 2026
Blocks per day	6	12
Items per block	38-39	18-20
Minutes per block	60	30
Total Day 1 items	232	232
Approx Day 1 length	~7 hours	~7 hours
Return to closed block	Not allowed	Not allowed

Per-item pacing target sits at roughly 90 seconds in both eras (60 minutes / 38-39 items and 30 minutes / 18-20 items are both around the 90 second mark). The pacing change is not in the average per-item budget. It is in the variance budget within a single block. A 3-minute spend on one item used to be absorbed across 37 other items inside the same block, now it has to be absorbed across 17-19. The recovery tax on a slow stem roughly doubles.

Day 2 MCQ: block container before and after

Day 2 has both the MCQ blocks and the CCS cases. The MCQ block structure shifted in the same direction Day 1 did, but the math is slightly different because the old Day 2 MCQ block was 45 minutes, not 60. Total MCQ item count holds at 180.

Feature	Through March 9, 2026	From March 10, 2026
MCQ blocks per day	6	9
Items per MCQ block	30	20
Minutes per MCQ block	45	30
Total Day 2 MCQ items	180	180
Approx Day 2 length	~9 hours	~9 hours
Return to closed block	Not allowed	Not allowed

The Day 2 MCQ per-item target was 90 seconds before (45 min / 30 items) and is 90 seconds after (30 min / 20 items). Again, the average per-item budget did not change. The variance budget did. A slow item used to spread across 29 other items in the same block, now it spreads across 19.

The other format shift: paragraph vignette vs chart-tabular

A Step 3 single-item or sequential-item-set patient scenario can appear in one of two surface layouts. The clinical content is identical between them. The reasoning task is identical. What changes is how the case is delivered to your eyes.

Paragraph vignette. A narrative block of text, typically 70-180 words. Demographic opener (age plus relevant context), presentation (chief complaint plus duration), exam-labs-imaging (the 1-3 findings the case turns on), occasional mechanism bridge, then the stem. Reads like a case report.

Chart-tabular layout. The same facts, but organized in clearly marked sections: Patient Information, History of Present Illness, Past Medical History, Social History, Medications, Allergies, Physical Exam, and Labs / Imaging. Familiar medical abbreviations are used. Reads like the structured note section of an EHR.

Step 1 and Step 2 CK do not do this. Their items are paragraph vignettes only. Step 3 is the exam that uniquely toggles between the two layouts inside the same block. Most prep articles about "Step 3 format" cover one or the other, rarely both. Both can appear in the same block on exam day, and your eye has to handle both.

The same case, both layouts, side by side

One patient scenario, both surface forms. The clinical facts are identical, the right answer is identical, the distractors are identical. What differs is how your eye walks the page. Read both and notice which one you finish first.

Paragraph vignette

A 58-year-old man comes to the clinic for a 3-month follow-up. He has a history of type 2 diabetes mellitus, hypertension, and hyperlipidemia. Current medications are metformin 1000 mg twice daily, lisinopril 20 mg daily, and atorvastatin 40 mg daily. He smokes one pack of cigarettes per day and drinks two beers most evenings. He reports no chest pain, shortness of breath, or claudication. Vital signs show blood pressure 148/92 mm Hg, pulse 84/min, and BMI 31 kg/m². Physical examination shows bilateral ankle edema, 1+. Laboratory studies show hemoglobin A1c 8.4%, LDL 142 mg/dL, eGFR 58 mL/min/1.73 m², and urine albumin-to-creatinine ratio 78 mg/g.

Chart-tabular layout

Patient Information: 58-year-old man, 3-month follow-up visit.
Past Medical History: T2DM, HTN, hyperlipidemia.
Medications: Metformin 1000 mg BID, lisinopril 20 mg daily, atorvastatin 40 mg daily.
Social History: 1 PPD smoker, ~2 beers most evenings.
Vitals / Exam: BP 148/92, HR 84, BMI 31. No CP, SOB, claudication. 1+ bilateral ankle edema.
Labs: A1c 8.4%, LDL 142, eGFR 58, urine ACR 78 mg/g.

Stem (identical for both layouts)

Which of the following is the most appropriate next step in management?

Add amlodipine 5 mg daily
Add empagliflozin 10 mg daily
Increase atorvastatin to 80 mg daily
Refer for renal biopsy
Repeat urine albumin-to-creatinine ratio in 6 months

Illustrative example written for layout comparison, not a released NBME item. The reasoning task (diabetic with stage 3a CKD, albuminuria, suboptimal A1c, suboptimal BP) is the same in both. The chart layout makes the labs cluster and the meds cluster easier to scan for the SGLT2-eligible flags; the paragraph layout reads more like a clinic note. Most examinees favor one parse rhythm over the other. The drill below closes the gap.

Anchor fact · why the chart-vs-paragraph delta is now load-bearing

A 30-second parse penalty cost 0.8s/item before. Now it costs 1.7s/item.

Under the old Day 1 container (60 minutes, 38-39 items), a 30-second parse-time penalty on one slow-layout item had to be recovered across the other 37 items in the block. That is roughly 0.8 seconds of additional pressure per recovery item. Manageable. Most examinees never noticed which layout they were reading.

Under the new Day 1 container (30 minutes, 18-20 items), the same 30-second penalty has to be recovered across 17-19 items. That is roughly 1.7 seconds of additional pressure per recovery item, slightly more than double the old tax. The chart-vs-paragraph parse-speed gap that did not previously rise above the noise floor now sits inside the range that matters.

The practical implication: if there is a layout type that you read 10-15 seconds slower than the other on average, it is worth closing that gap before exam day. The cheapest way to close it is timed drilling on the slower layout in isolation until parse speed converges with the faster one.

A five-step drill for the layout gap

This is the practice loop for the chart-vs-paragraph parse-time gap, layered on top of the standard timed-block work you should already be doing under the new container. Run it over 1-2 weeks during dedicated, in parallel with your normal QBank rotation.

Closing the chart-vs-paragraph parse-time gap

Pull 20 items where layout is your call

In your QBank, build a custom 20-item Step 3 block. Do not pre-filter for layout. You want a real mix of paragraph vignettes and chart-tabular items, the way exam day will deliver them.

Run a cold 30-minute timed block

External timer. No pauses, no lookups. Mark and skip inside the block is fine. Note which items felt slow to parse before you committed to an answer.

Tag each missed or slow item by layout

Was it a paragraph vignette or a chart-tabular item? Track this in a notes file across 4-6 timed blocks. If a single layout type is responsible for most of your time pressure, that is the one to drill in isolation.

Drill the slow layout in 20-item-only sets

Most QBanks let you filter by media type or by question type, but not always by layout. AMBOSS exposes chart-format items more explicitly than UWorld. Spend two or three custom blocks doing only that layout cold, with a 30-minute timer, until parse speed converges.

Drop back to mixed-layout timed blocks

Once the gap closes, run 30-minute, 20-item mixed blocks again. The point is not to be faster on one layout, it is to lose your sensitivity to the difference. By exam day you want the layout to feel like noise, not signal.

The under-practiced piece is step 3, tagging missed or slow items by layout. Most examinees track only the content domain (cards, renal, OB, ethics) and miss that there is a structural variable orthogonal to the content. If your tag log shows you are slow on chart-tabular items specifically, you have a fix-able mechanical gap that costs you nothing on content knowledge and roughly 5-10 seconds per item on parse.

Where the between-block five-minute drill fits

UWorld and AMBOSS remain the right tools for the full 30-minute, 20-item timed blocks. The question is what to do on the days you do not have time or attention for two or three full timed blocks, but you still have access to a residency teaching deck, a board-review chapter, or a recorded lecture.

One Studyly drill session is about 5 minutes of about 5 questions. A real new-format Step 3 block is 30 minutes of 20 questions. So one Studyly session is roughly one-sixth of a real block by time, one-quarter by item count. Four back-to-back five-minute sessions on the same deck is the closest informal-timer practice you can get to one real new-format block, minus the no-return-to-closed rule.

The case-style generator produces roughly 60-80 vignette-format cards from a 90-slide deck in about 60 seconds. The vignette template matches the paragraph layout an NBME item uses, covering the chart-tabular gap by being explicitly paragraph, which is the layout the chart-tabular reader needs more reps on. The cards are anchored to specific slide numbers from your source deck, so the explain panel on a miss takes you back to the slide you originally read.

The four-criterion question-quality eval, on a held-out three-document set graded for factual correctness, clarity, distractor quality, and question-type coverage:

0Studyly

0Unattle

0Gauntlet

0Turbolearn

Higher is better. Three source documents (a slide deck, a textbook chapter, a paper) held out. Each tool generated questions from the same three documents. Every output graded on the same rubric by the same graders. Methodology and per-document sub-scores are on the quality page.

The mistake worth not making

The common mistake on the Step 3 format shift is to read about the March 10 block change, conclude that nothing changes about content, and then change nothing at all about how you drill. That is half right. Content did not change, organ-system weighting did not change, vignette length did not change. But the container did, and the container shapes how you spend the 90-second per-item budget. A study plan that ignores the variance-budget compression inside a block leaves performance on the table that costs nothing on content to recover.

The opposite mistake is to panic and rebuild a study plan from scratch. Do not. Run the same content schedule, switch your timed-block size from 60 (or 45) minutes to 30, and add a layout-tagging pass on the first 4-6 timed blocks you do under the new format. That covers both shifts with a small mechanical change on top of work you would have done anyway.

For the interface itself, USMLE provides the testing experience on the official update page. Spend 10-15 minutes there before your exam date. Learn where the settings menu lives, what the per-image contrast slider does, and never think about it again.

Drill between timed blocks

Drop a residency teaching deck. Get 60-80 vignette cards in 60 seconds.

Free tier on app.jungleai.com, no card. Case-style cards in paragraph-vignette format, anchored to specific slide numbers from your source. Works on PowerPoint, PDF, Keynote, scanned slides, textbook chapters, and YouTube lectures.

Common questions about the Step 3 format shift

What is the Step 3 format shift in 2026?

On March 10, 2026, Step 3 moved to a new test delivery software platform. The structural change is the block container, not the items inside it. Day 1 went from 6 blocks of 38-39 items (60 minutes each) to 12 blocks of 18-20 items (30 minutes each). Day 2 MCQ went from 6 blocks of 30 items (45 minutes each) to 9 blocks of 20 items (30 minutes each). Total item counts stay at 232 for Day 1 and 180 for Day 2 MCQ. Total exam day length stays approximately 7 hours for Day 1 and 9 hours for Day 2. The pass standard does not change. The rule that you cannot return to a closed block remains in effect. Verified against the official USMLE test delivery software updates page on 2026-05-20.

Did the vignette format itself change?

No. The item structure, the clinical vignette template (demographic plus presentation plus exam-labs-imaging plus optional mechanism bridge plus stem), the length of stems, the answer choice format, the sequential item set rules, and the use of media items are all unchanged. What changed is the interface around the items and the block timer. A vignette that read like a UWorld Step 3 stem on March 9 still reads like that stem on March 10. The parse work inside the head is identical. The clock around it is different.

What is the chart-tabular vs paragraph vignette format on Step 3?

Step 3 is the only USMLE exam where a single patient scenario can render in two different layouts: as a traditional paragraph vignette, or as a chart-tabular layout that organizes the same facts into clearly marked sections (Patient Information, History of Present Illness, Past Medical History, Social History, Medications, Allergies, Physical Exam, Labs/Imaging). The chart format mirrors the structure of an actual EHR note. The same clinical scenario, the same answer, the same distractors, just a different surface. Step 1 and Step 2 CK do not present items in chart-tabular form. So when people say 'Step 3 format shift', it can mean two completely different things, and which one matters depends on what they are asking about.

Why does the chart-vs-paragraph distinction matter more under the new 30-minute blocks?

Because the variance budget per block has compressed. Under the old 60-minute Day 1 block of 38-39 items, a 30-second parse-time penalty on one item could be absorbed across the other 37 items, costing roughly 0.8 seconds per recovery item. Under the new 30-minute block of 18-20 items, that same 30-second penalty has to be absorbed across 17-19 items, costing roughly 1.7 seconds per recovery item. Roughly doubled. So if you are habitually slow on one of the two layouts, the new container charges you a bigger tax for that slowness than the old container did. Practicing both layouts cold is now load-bearing in a way it was not before March 10.

Which format is faster, vignette or chart?

It depends on which one your residency reading habits match. Residents who spend their day reading actual EHR notes (where labs sit in a section, meds sit in a section, exam findings sit in a section) often parse chart-tabular items faster because the layout matches what their eyes are used to. Residents whose reading habits skew toward narrative case write-ups often parse paragraph vignettes faster because the prose flows. There is no published timing data on average parse-time delta between the two formats. The practical move is to time yourself on a mixed block of both layouts, on the same source material, and see which one your eyes default to. Then drill the other one until they converge.

How should daily practice change for the new Step 3 block format?

Drill in 30-minute, 20-item blocks under an external timer. Do not pause mid-block. Do not look anything up mid-block. Mark and skip is allowed inside the block; lookups are not. Close the block on the buzzer, regardless of unanswered items, then review missed items for 10-15 minutes. Two to four blocks back-to-back is a reasonable session. The block count doubled for Day 1 and grew by 50 percent for Day 2 MCQ, which means more closes per exam day. The close discipline (commit before the buzzer, do not leave items blank) is now twice as load-bearing on Day 1 as it was on the old format.

What about CCS cases on Day 2?

CCS was restructured too. The case count dropped from 13 to 9 on Day 2, while the per-case time limit stayed at 25 minutes. Total CCS testing time falls from roughly 5.4 hours to 3.75 hours. The CCS interface itself was redesigned alongside the MCQ delivery software in the same March 10, 2026 rollout. Always confirm the current numbers on the official USMLE update page for the date you are sitting, since this is the area NBME has updated most quietly during prior software refreshes.

Will old UWorld and AMBOSS blocks still drill the right skill?

Yes. Item content is unchanged. UWorld and AMBOSS Step 3 banks built before March 10, 2026 still drill the right cognitive task. The only thing to change in how you use them is to build custom blocks of 20 items and set an external 30-minute timer rather than running their default block sizes. If a QBank does not natively expose chart-tabular layout, you can still drill the cognitive task on paragraph vignettes; the layout shift is mostly about parse speed, not about the underlying reasoning.

Where does Studyly fit into a Step 3 prep plan?

Studyly is not a replacement for UWorld or AMBOSS, which remain the standard commercial Step 3 QBanks for full timed-block drilling. The unique value is for your residency-specific material: upload a teaching-conference slide deck, a hospital protocol PDF, a board-review chapter, or a YouTube lecture, and get roughly 60-80 vignette-format multiple-choice items from a 90-slide deck in about 60 seconds. These are case-style cards in paragraph format, scored 81.3 on a held-out three-document eval. Use them for the five-minute drills between formal timed blocks. One Studyly drill session is roughly one-sixth of a real 30-minute Step 3 block by time. Four back-to-back five-minute drills on the same deck is the closest informal practice you can get to one new-format block.

Does the format shift change scoring?

No. USMLE has stated that the pass standard, the total item counts (232 for Day 1 MCQ, 180 for Day 2 MCQ), and the overall exam day length are unchanged. The exam remains scored on the existing standard with reporting on the three-digit scale. Items missed under the new container count the same as items missed under the old one. A correct answer on March 9 is worth what a correct answer on March 10 is worth.

Should I redo my study schedule because of the shift?

The high-level schedule does not need to change. The two adjustments are mechanical. First, change your timed-block practice from 60-minute (Day 1) or 45-minute (Day 2 MCQ) blocks to 30-minute, 20-item blocks. Second, deliberately drill on chart-tabular layout items in addition to paragraph vignettes, since the new tighter container makes parse-speed gaps more expensive than they were before. Both adjustments are about matching practice conditions to the new exam container. Content review, organ-system focus, and time-to-exam pacing are unchanged.

Paired guides on this site:

USMLE Step 1 new-format practice questions — the May 14, 2026 Step 1 block change, which mirrors the Step 3 shift but on different numbers (14 blocks of 20, not 12 of 18-20).
Step 1 to Step 2 CK prep timing — how the cross-step prep calendar shifts under the 2026 software updates.
USMLE vignette drilling from your own lecture decks — the five-minute between-block loop in detail, with the NBME-shaped item template the generator targets.
Distractor handling vs concept recall on USMLE items — why the four wrong answers are where Step 3 timing pressure actually lands, and how to drill that skill without burning a QBank attempt.