Qwen-Image-2.1-Turbo against its base model, Z-Image Turbo and FLUX.2 klein

Image modelsRun10 October 2026Updated10 October 2026OnRTX 5090, 32 GB, rented

Qwen-Image-2.1-Turbo, Portrait, seed 11 — 1Qwen-Image-2.1-Turbo

Qwen-Image-2.1, Portrait, seed 11 — 2Qwen-Image-2.1

Z-Image Turbo, Portrait, seed 11 — 3Z-Image Turbo

FLUX.2 klein 4B, Portrait, seed 11 — 4FLUX.2 klein 4B

Original overall verdict

Qwen-Image-2.1-Turbo ranked first in this AI-assisted blind review of 280 images, followed by Qwen-Image-2.1, Z-Image Turbo and FLUX.2 klein 4B. Z-Image Turbo led text rendering; Qwen-Image-2.1 led style and narrowly led anatomy. Every model made visible mistakes. Seven reviewers judged one round each with names hidden and labels shuffled per comparison. The overall order weights the seven rounds equally; speed and memory are separate. These subjective results apply to this prompt set, not a statistically validated universal ranking. Full review notes and limitations are linked below.

Reviewed comparison

Compare models

Review aspect-ranks-v2 · rubric aspect-ranks-v2 · 4 contestants · 70 reviewed comparisons

7 independent AI reviewers · 70 reviewed comparisons.

Review method and reviewer provenance

Five independent aspect rankings per prompt and seed, with randomized blind labels and full prompts. One independent AI reviewer per round inspected original images. Ties share their mean ordinal position. Equal seed weight within prompt, equal prompt weight within round, equal round weight overall. Task performance combines adherence and correctness equally; Taste combines visual quality, composition and aesthetic appeal equally. Composite charts use common complete coverage. These are relative, subjective rankings for this contestant lineup, not absolute quality ratings or general intelligence. Time and estimated execution cost were withheld from reviewers. Cost uses the recorded rental estimate and excludes setup, idle time and bandwidth; it is not an API price or final bill.

Reviewer instances: gpt-6-astra / independent round reviewer / alpha round, gpt-6-astra / independent round reviewer / anatomy round, gpt-6-astra / independent round reviewer / following round, gpt-6-astra / independent round reviewer / hires round, gpt-6-astra / independent round reviewer / realism round, gpt-6-astra / independent round reviewer / style round, gpt-6-astra / independent round reviewer / text round.

Mean rank

Compare with

Resolution

Start state

Set winners have the lowest mean rank across all five aspects, weighted equally; equal means tie. Lower mean rank is better. Tied contestants share the mean of their occupied places. Task performance averages Prompt adherence and Correctness within each comparison; seeds, prompts and rounds receive equal weight at their level.

Task performance

70 complete comparison sets used from 70. 0 failed images were not ranked.

Qwen-Image-2.1-Turbo2.02

Qwen-Image-2.12.3

Z-Image Turbo2.83

FLUX.2 klein 4B2.84

Task performance mean ranks. Lower is better.
Contestant	Mean rank	Complete comparisons
Qwen-Image-2.1-Turbo	2.02	70
Qwen-Image-2.1	2.3	70
Z-Image Turbo	2.83	70
FLUX.2 klein 4B	2.84	70

Task performance and generation time

Warm results only · at most 1.6 MP · 65 matched comparison sets.

Comparable results. Median seconds per image; lower mean rank is better.
Contestant	Mean rank	Median seconds per image	Matched comparisons
Qwen-Image-2.1-Turbo	2.12	3.75 s	65
Qwen-Image-2.1	2.41	16.12 s	65
Z-Image Turbo	2.65	2.82 s	65
FLUX.2 klein 4B	2.82	1.21 s	65

Excluded: 3 other start state, 2 other resolution class.

Measured on the same machine

Seconds per image at about 1 megapixel, model loaded, s

Lower is better

Qwen-Image-2.1-Turbo3.75

Qwen-Image-2.116.12

Z-Image Turbo2.82

FLUX.2 klein 4B1.21

Peak whole-device graphics memory in use, GB

Lower is better

Qwen-Image-2.1-Turbo30.9

Qwen-Image-2.130.9

Z-Image Turbo30.8

FLUX.2 klein 4B26.3

How it was run

Results made: 280
Shown here: All of them
Recorded cost: $0.75Provider-reported rental subtotal as of 10 October 2026, 14:15 UTC. The rental was still running; this is not the final bill and excludes AI review costs.
Versions and settings: Per contestant →

1 of 7

Realism

WinnerQwen-Image-2.1-Turbo. Qwen-Image-2.1-Turbo led the blind review across portraits, hands and street scenes. Portrait differences were often close. Even the stronger hand images sometimes had extra hands, unclear finger connections or an implausible card shuffle.

Portrait

Close-up portrait photo of a 60-year-old fisherman with weathered skin, grey stubble and pale blue eyes, wearing a wet yellow raincoat, overcast daylight on a harbor, 85mm lens, shallow depth of field, natural skin texture with visible pores

Review reasons

These comparison reasons apply to the set as a whole.

Comparison reason · Prompt adherence: Blind labels: A = Qwen-Image-2.1-Turbo; B = Z-Image Turbo; C = Qwen-Image-2.1; D = FLUX.2 klein 4B. B makes wet droplets and a fishing harbor clearest; A/C supply stronger close-up texture and shallow focus than D's smoother face.Confidence: medium
Comparison reason · Correctness: Blind labels: A = Qwen-Image-2.1-Turbo; B = Z-Image Turbo; C = Qwen-Image-2.1; D = FLUX.2 klein 4B. All four have plausible facial anatomy and garment structure. D can be wearing a separate yellow rain cap above a lowered hood; no impossible connection is visible, so this does not support a correctness penalty.Confidence: high
Comparison reason · Visual quality: Blind labels: A = Qwen-Image-2.1-Turbo; B = Z-Image Turbo; C = Qwen-Image-2.1; D = FLUX.2 klein 4B. A/C resolve fine hair and pores cleanly; B's cheek texture is more lumpy, and D's stubble is bright and wiry against smoother skin.Confidence: medium
Comparison reason · Composition: Blind labels: A = Qwen-Image-2.1-Turbo; B = Z-Image Turbo; C = Qwen-Image-2.1; D = FLUX.2 klein 4B. A/C fill the frame with the eyes and face while keeping context soft; B's boats compete slightly, and D allocates more area to coat.Confidence: medium
Comparison reason · Aesthetic appeal: Blind labels: A = Qwen-Image-2.1-Turbo; B = Z-Image Turbo; C = Qwen-Image-2.1; D = FLUX.2 klein 4B. A/C have cohesive muted harbor color and expressive natural detail; B is compelling but harsher, D feels more polished and uniform.Confidence: medium

Hands

Photo of two hands shuffling a deck of playing cards over a green felt table, cards arcing between the hands in mid-air, warm lamp light, sharp focus on the fingers

Qwen-Image-2.1, Hands, seed 11 — *Qwen-Image-2.1*

Z-Image Turbo, Hands, seed 11 — *Z-Image Turbo*

FLUX.2 klein 4B, Hands, seed 11 — *FLUX.2 klein 4B*

Review reasons

These comparison reasons apply to the set as a whole.

Comparison reason · Prompt adherence: Blind labels: A = FLUX.2 klein 4B; B = Z-Image Turbo; C = Qwen-Image-2.1-Turbo; D = Qwen-Image-2.1. C/D show two sharp hands manipulating an arched deck under warm light, though the cards remain bunched; B has just one airborne card, and A adds two extra hands.Confidence: high
Comparison reason · Correctness: Blind labels: A = FLUX.2 klein 4B; B = Z-Image Turbo; C = Qwen-Image-2.1-Turbo; D = Qwen-Image-2.1. C has the most coherent grip despite malformed card indices; D's cards fuse into irregular layers, B's airborne A card carries face-card graphics, and A has four hands and mixed suits.Confidence: medium
Comparison reason · Visual quality: Blind labels: A = FLUX.2 klein 4B; B = Z-Image Turbo; C = Qwen-Image-2.1-Turbo; D = Qwen-Image-2.1. C/D offer crisp skin and felt with consistent warm light; B is clean but softer in cards, while A's scattered card shapes and overlapping hands are less controlled.Confidence: medium
Comparison reason · Composition: Blind labels: A = FLUX.2 klein 4B; B = Z-Image Turbo; C = Qwen-Image-2.1-Turbo; D = Qwen-Image-2.1. C presents the action cleanly between balanced hands; D is tighter and busier, B's lamp and isolated card pull attention upward, A splits focus among four hands.Confidence: high
Comparison reason · Aesthetic appeal: Blind labels: A = FLUX.2 klein 4B; B = Z-Image Turbo; C = Qwen-Image-2.1-Turbo; D = Qwen-Image-2.1. C's intimate warm palette and simple focus are most cohesive; D is similar but cramped, B feels staged, A feels chaotic.Confidence: medium

Street

Candid street photo of a crowded night market in the rain, people with umbrellas, food stalls with rising steam, neon signs reflected in puddles, shot on 35mm film

Qwen-Image-2.1, Street, seed 11 — *Qwen-Image-2.1*

Z-Image Turbo, Street, seed 11 — *Z-Image Turbo*

FLUX.2 klein 4B, Street, seed 11 — *FLUX.2 klein 4B*Winner

Review reasons

These comparison reasons apply to the set as a whole.

Comparison reason · Prompt adherence: Blind labels: A = Qwen-Image-2.1; B = Z-Image Turbo; C = Qwen-Image-2.1-Turbo; D = FLUX.2 klein 4B. D most fully combines night, dense umbrellas, steaming food, neon reflections and film grain. A/C have convincing night-market atmosphere but weaker neon; B's bright blue sky reads as twilight.Confidence: high
Comparison reason · Correctness: Blind labels: A = Qwen-Image-2.1; B = Z-Image Turbo; C = Qwen-Image-2.1-Turbo; D = FLUX.2 klein 4B. B/D show clear body, umbrella and stall relationships; C's small stall workers are less resolved, and A's left-center figures and tilted umbrella overlap into indistinct shapes.Confidence: medium
Comparison reason · Visual quality: Blind labels: A = Qwen-Image-2.1; B = Z-Image Turbo; C = Qwen-Image-2.1-Turbo; D = FLUX.2 klein 4B. B retains clean wet-surface and face detail; D's grain is coherent with film while keeping silhouettes clear. C loses more shadow detail; A has more smeared people and bright stall edges.Confidence: medium
Comparison reason · Composition: Blind labels: A = Qwen-Image-2.1; B = Z-Image Turbo; C = Qwen-Image-2.1-Turbo; D = FLUX.2 klein 4B. D's signs, umbrella rhythm and reflections guide the eye down the market; B provides strong foreground subjects, C is darker and flatter, and A's foreground scooter competes with the crowd.Confidence: medium
Comparison reason · Aesthetic appeal: Blind labels: A = Qwen-Image-2.1; B = Z-Image Turbo; C = Qwen-Image-2.1-Turbo; D = FLUX.2 klein 4B. D's pastel neon and film texture are cohesive; C has a convincing restrained nocturnal mood, B is appealing but more polished, A is visually muddier.Confidence: medium

2 of 7

Text rendering

WinnerZ-Image Turbo. Z-Image Turbo ranked first across these poster, banking-screen and shop-sign sets. It more consistently preserved the requested words and numbers. Small print, punctuation and Japanese lettering remained imperfect or hard to judge in some images.

Poster

A screen-printed concert poster in two colors, orange and navy. Large headline at the top reads "MIDNIGHT STATIC". Below an illustration of a radio tower, a subline reads "Live at the Old Cannery — Saturday 14 November". Small print along the bottom edge reads "Doors 7 PM · Tickets $25 · All ages".

Qwen-Image-2.1, Poster, seed 11 — *Qwen-Image-2.1*

Z-Image Turbo, Poster, seed 11 — *Z-Image Turbo*Winner

FLUX.2 klein 4B, Poster, seed 11 — *FLUX.2 klein 4B*

Review reasons

These comparison reasons apply to the set as a whole.

Comparison reason · Prompt adherence: Blind labels: A = FLUX.2 klein 4B; B = Qwen-Image-2.1; C = Qwen-Image-2.1-Turbo; D = Z-Image Turbo. D preserves all content and places the subline beneath the tower. B/C put it above; A corrupts the date/venue and footer.Confidence: high
Comparison reason · Correctness: Blind labels: A = FLUX.2 klein 4B; B = Qwen-Image-2.1; C = Qwen-Image-2.1-Turbo; D = Z-Image Turbo. B/C/D reproduce the requested wording and $25. A has 'Dooors', a garbled subline and extra characters around the price.Confidence: high
Comparison reason · Visual quality: Blind labels: A = FLUX.2 klein 4B; B = Qwen-Image-2.1; C = Qwen-Image-2.1-Turbo; D = Z-Image Turbo. D has clean controlled print texture and coherent simplified lines. C has muddier fine lattice; B more clogged tower detail. A malforms lower letters.Confidence: medium
Comparison reason · Composition: Blind labels: A = FLUX.2 klein 4B; B = Qwen-Image-2.1; C = Qwen-Image-2.1-Turbo; D = Z-Image Turbo. D gives every information tier room. B has dynamic tower perspective and clear footer. C has a less active central field. A crowds lower text and touches title/tower.Confidence: medium
Comparison reason · Aesthetic appeal: Blind labels: A = FLUX.2 klein 4B; B = Qwen-Image-2.1; C = Qwen-Image-2.1-Turbo; D = Z-Image Turbo. B's radiating background and tower angle create concert energy. D is cohesive and restrained; C less dynamic. A's inconsistent lower type weakens its appeal.Confidence: medium

App screen

A mobile banking app screen, clean modern flat UI on a light background. Header text "Good morning, Maya". A balance card showing "$12,480.55". Three round buttons labelled "Send", "Request" and "Top up". A list titled "Recent activity" with three rows: "Coffee Lab −$4.50", "Salary +$3,200.00", "Rent −$1,150.00".

Qwen-Image-2.1, App screen, seed 11 — *Qwen-Image-2.1*Winner

Z-Image Turbo, App screen, seed 11 — *Z-Image Turbo*

FLUX.2 klein 4B, App screen, seed 11 — *FLUX.2 klein 4B*

Review reasons

These comparison reasons apply to the set as a whole.

Comparison reason · Prompt adherence: Blind labels: A = Z-Image Turbo; B = Qwen-Image-2.1; C = FLUX.2 klein 4B; D = Qwen-Image-2.1-Turbo. B/D give light flat screens, balance cards and complete rows, but pills. A has circles yet adds a rendered phone, raised controls and lacks a balance card. C adds a device and corrupts amounts/rows.Confidence: high
Comparison reason · Correctness: Blind labels: A = Z-Image Turbo; B = Qwen-Image-2.1; C = FLUX.2 klein 4B; D = Qwen-Image-2.1-Turbo. A/B/D preserve all required text and amounts. C shows Salary +$3,2000.00 and Rent -$1150.0, with the rent amount on another line.Confidence: high
Comparison reason · Visual quality: Blind labels: A = Z-Image Turbo; B = Qwen-Image-2.1; C = FLUX.2 klein 4B; D = Qwen-Image-2.1-Turbo. B has consistent type/separators. A is crisply rendered despite off-brief depth. D's icons are scrappy and inconsistent. C has unstable type and clipped layout.Confidence: medium
Comparison reason · Composition: Blind labels: A = Z-Image Turbo; B = Qwen-Image-2.1; C = FLUX.2 klein 4B; D = Qwen-Image-2.1-Turbo. D separates balance/activity and aligns amounts. B leaves more unused upper-card space. A spends canvas on phone framing; C breaks and clips its final row.Confidence: high
Comparison reason · Aesthetic appeal: Blind labels: A = Z-Image Turbo; B = Qwen-Image-2.1; C = FLUX.2 klein 4B; D = Qwen-Image-2.1-Turbo. B is a cohesive restrained blue-white UI. D is orderly but icons vary. A is polished yet less suited to flat UI. C looks unfinished.Confidence: medium

Shop sign

Photo of a small ramen shop front at dusk. A wooden sign above the door reads "TANUKI RAMEN" with the Japanese text "たぬきラーメン" beneath it. A chalkboard beside the door says "Open 11–22 · Cash only".

Qwen-Image-2.1, Shop sign, seed 11 — *Qwen-Image-2.1*Winner

Z-Image Turbo, Shop sign, seed 11 — *Z-Image Turbo*

FLUX.2 klein 4B, Shop sign, seed 11 — *FLUX.2 klein 4B*

Review reasons

These comparison reasons apply to the set as a whole.

Comparison reason · Prompt adherence: Blind labels: A = FLUX.2 klein 4B; B = Qwen-Image-2.1-Turbo; C = Qwen-Image-2.1; D = Z-Image Turbo. D places a complete bilingual wood sign above the door and a readable chalkboard beside it. C/B split RAMEN away from TANUKI and omit the chalkboard dot; B's Japanese ending is degraded. A crops the board and adds repeated/garbled cash wording.Confidence: medium
Comparison reason · Correctness: Blind labels: A = FLUX.2 klein 4B; B = Qwen-Image-2.1-Turbo; C = Qwen-Image-2.1; D = Z-Image Turbo. D clearly preserves the required text. C's main wording reads correctly but lacks the requested chalkboard separator. B's Japanese ending is visibly degraded. A's board reads 'Cash ony' plus a repeated 'only'; its second kana is uncertain.Confidence: medium
Comparison reason · Visual quality: Blind labels: A = FLUX.2 klein 4B; B = Qwen-Image-2.1-Turbo; C = Qwen-Image-2.1; D = Z-Image Turbo. C has convincing weathered wood, glass and dusk lighting. B is similarly coherent with slightly less controlled sign lettering. D is clean but has smoother, more synthetic materials. A has more distorted incidental text and board lettering.Confidence: medium
Comparison reason · Composition: Blind labels: A = FLUX.2 klein 4B; B = Qwen-Image-2.1-Turbo; C = Qwen-Image-2.1; D = Z-Image Turbo. D prioritizes both required signs without cropping. C keeps the entire board and a balanced entrance. B's board meets the lower edge. A crops the board and gives the shop a more awkward off-center framing.Confidence: medium
Comparison reason · Aesthetic appeal: Blind labels: A = FLUX.2 klein 4B; B = Qwen-Image-2.1-Turbo; C = Qwen-Image-2.1; D = Z-Image Turbo. C's weathered facade and localized amber light feel cohesive. B has appealing lantern warmth. D is welcoming but more showroom-like; A is busier and its cut-off board weakens the image.Confidence: medium

3 of 7

Cartoon and style

WinnerQwen-Image-2.1. Qwen-Image-2.1 led this mix of 3D illustration, pixel art and anime. Results were judged against the requested style rather than photographic realism; several images still missed scene details or used an inconsistent pixel grid.

Anime

1990s cel-animation anime screenshot: a girl with short red hair riding a bicycle down a seaside hill, towering summer clouds, flat cel shading, clean line art, slight film grain

Qwen-Image-2.1, Anime, seed 11 — *Qwen-Image-2.1*

Z-Image Turbo, Anime, seed 11 — *Z-Image Turbo*

FLUX.2 klein 4B, Anime, seed 11 — *FLUX.2 klein 4B*Winner

Review reasons

These comparison reasons apply to the set as a whole.

Comparison reason · Prompt adherence: Blind labels: A = FLUX.2 klein 4B; B = Qwen-Image-2.1-Turbo; C = Z-Image Turbo; D = Qwen-Image-2.1. A most explicitly combines flat cel planes, short red hair, large clouds and a sloping coastal road. B/D retain convincing period cel characters but softer painted backgrounds; C's road reads almost level.Confidence: medium
Comparison reason · Correctness: Blind labels: A = FLUX.2 klein 4B; B = Qwen-Image-2.1-Turbo; C = Z-Image Turbo; D = Qwen-Image-2.1. B has the clearest connected riding pose and bicycle controls. A is largely coherent but pedal/foot layering is less clear. C has a conspicuously isolated-looking far shoe below the crank; D's front wheel spokes/hub are poorly resolved.Confidence: medium
Comparison reason · Visual quality: Blind labels: A = FLUX.2 klein 4B; B = Qwen-Image-2.1-Turbo; C = Z-Image Turbo; D = Qwen-Image-2.1. A sustains clean contours and controlled flat shading throughout. B/D have comparably coherent cel characters over painted scenery. C's cloud and foreground marks are somewhat blotchier, while its bicycle linework is less resolved.Confidence: medium
Comparison reason · Composition: Blind labels: A = FLUX.2 klein 4B; B = Qwen-Image-2.1-Turbo; C = Z-Image Turbo; D = Qwen-Image-2.1. A/C frame the entire bicycle with strong cloud/sea hierarchy. D also balances well but crowds the bicycle lower edge; B cuts the front wheel and gives the rider less breathing room.Confidence: medium
Comparison reason · Aesthetic appeal: Blind labels: A = FLUX.2 klein 4B; B = Qwen-Image-2.1-Turbo; C = Z-Image Turbo; D = Qwen-Image-2.1. B/D's subdued teal skies and hand-painted coastal textures create a cohesive nostalgic mood. A/C are appealing alternatives, with more graphic clouds in A and brighter painted shapes in C; this is a close stylistic preference.Confidence: low

3D film still

Still from a 3D animated family film: a small round robot gardener watering a sunflower twice its size inside a sunlit greenhouse, soft global illumination, big expressive eyes

Qwen-Image-2.1, 3D film still, seed 11 — *Qwen-Image-2.1*

Z-Image Turbo, 3D film still, seed 11 — *Z-Image Turbo*

FLUX.2 klein 4B, 3D film still, seed 11 — *FLUX.2 klein 4B*

Review reasons

These comparison reasons apply to the set as a whole.

Comparison reason · Prompt adherence: Blind labels: A = Qwen-Image-2.1; B = FLUX.2 klein 4B; C = Z-Image Turbo; D = Qwen-Image-2.1-Turbo. D most closely gives a round expressive gardener, roughly twice-height sunflower and water reaching its roots. A performs the action well but the flower is less oversized. B has two competing flowers and much water missing the pot; C leaks beside the plant and the height contrast is smaller.Confidence: high
Comparison reason · Correctness: Blind labels: A = Qwen-Image-2.1; B = FLUX.2 klein 4B; C = Z-Image Turbo; D = Qwen-Image-2.1-Turbo. A/D show coherent robots and water directed into the planted soil. B's downward side stream lands outside the pot. C visibly drains from the bottom of its open cup into a separate puddle, an incoherent pouring mechanism.Confidence: high
Comparison reason · Visual quality: Blind labels: A = Qwen-Image-2.1; B = FLUX.2 klein 4B; C = Z-Image Turbo; D = Qwen-Image-2.1-Turbo. A/D have clean articulated forms, convincing soft bounce light and controlled water/material detail. B is comparably finished but the split water streams are less coherent. C's smooth toy surfaces work stylistically, but its droplets and flat ring puddle feel less integrated.Confidence: medium
Comparison reason · Composition: Blind labels: A = Qwen-Image-2.1; B = FLUX.2 klein 4B; C = Z-Image Turbo; D = Qwen-Image-2.1-Turbo. D gives the clearest flower-robot-action triangle. A is similarly strong but its held stem and can overlap more. C has a readable simple pairing; B's second large flower competes with the watering subject.Confidence: high
Comparison reason · Aesthetic appeal: Blind labels: A = Qwen-Image-2.1; B = FLUX.2 klein 4B; C = Z-Image Turbo; D = Qwen-Image-2.1-Turbo. D/A's expressive faces, warm light and tactile gardener design are equally appealing. C's gentle simplified robot is charming despite its water problem. B's two-flower arrangement and less expressive face make the interaction less engaging.Confidence: low

Pixel art

16-bit pixel art, side-scrolling game scene: a knight standing before a castle gate at sunset, limited color palette, crisp square pixels, no anti-aliasing

Qwen-Image-2.1, Pixel art, seed 11 — *Qwen-Image-2.1*Winner

Z-Image Turbo, Pixel art, seed 11 — *Z-Image Turbo*

FLUX.2 klein 4B, Pixel art, seed 11 — *FLUX.2 klein 4B*

Review reasons

These comparison reasons apply to the set as a whole.

Comparison reason · Prompt adherence: Blind labels: A = FLUX.2 klein 4B; B = Z-Image Turbo; C = Qwen-Image-2.1; D = Qwen-Image-2.1-Turbo. B has the strongest palette restraint and block silhouettes; C/D have more convincing 16-bit sprite treatment and side-scrolling ground. Their different strengths are close. A reads more like a perspective illustration with fine smooth contours than a side-scrolling pixel scene.Confidence: medium
Comparison reason · Correctness: Blind labels: A = FLUX.2 klein 4B; B = Z-Image Turbo; C = Qwen-Image-2.1; D = Qwen-Image-2.1-Turbo. All show one coherent knight standing before a readable castle entrance at sunset. Simplified hands and architecture do not establish a meaningful objective correctness difference.Confidence: medium
Comparison reason · Visual quality: Blind labels: A = FLUX.2 klein 4B; B = Z-Image Turbo; C = Qwen-Image-2.1; D = Qwen-Image-2.1-Turbo. B's large blocks and C's smaller sprite clusters are comparably readable within pixel art. D loses some castle/ground definition to soft tonal shading. A's thin near-smooth outlines and mixed pixel scale are less consistent with crisp-square execution.Confidence: medium
Comparison reason · Composition: Blind labels: A = FLUX.2 klein 4B; B = Z-Image Turbo; C = Qwen-Image-2.1; D = Qwen-Image-2.1-Turbo. C/D balance the left knight and castle against the right sunset while preserving a continuous gameplay plane. B's symmetric gate frames its knight strongly but is more static. A's perspective courtyard weakens the side-scrolling organization.Confidence: medium
Comparison reason · Aesthetic appeal: Blind labels: A = FLUX.2 klein 4B; B = Z-Image Turbo; C = Qwen-Image-2.1; D = Qwen-Image-2.1-Turbo. C/D's warm sunset palettes and readable armored sprites are comparably cohesive. B has appealing graphic restraint but less individual character. A is attractive illustration, yet its medium feels less internally consistent for this request.Confidence: low

4 of 7

Prompt following

WinnerQwen-Image-2.1-Turbo. Qwen-Image-2.1-Turbo led the object-count and spatial-instruction sets, with FLUX.2 klein next. Cup counts and the three-person pose exposed mistakes across models. The wording of left and right in the group prompt was ambiguous, so those judgments are qualified.

Cups and teapot

Three red cups in a row to the left of one blue teapot on a wooden table, a single yellow lemon in front of the teapot, a green book lying flat at the right edge of the table, window light from the top left

Qwen-Image-2.1, Cups and teapot, seed 11 — *Qwen-Image-2.1*

Z-Image Turbo, Cups and teapot, seed 11 — *Z-Image Turbo*Winner

FLUX.2 klein 4B, Cups and teapot, seed 11 — *FLUX.2 klein 4B*

Review reasons

These comparison reasons apply to the set as a whole.

Comparison reason · Prompt adherence: Blind labels: A = Qwen-Image-2.1-Turbo; B = FLUX.2 klein 4B; C = Qwen-Image-2.1; D = Z-Image Turbo. All four show exactly three red cups left of one blue teapot, one foreground lemon, a flat green book at frame right and upper-left window light.Confidence: high
Comparison reason · Correctness: Blind labels: A = Qwen-Image-2.1-Turbo; B = FLUX.2 klein 4B; C = Qwen-Image-2.1; D = Z-Image Turbo. All have the correct object count and colors, coherent vessel connections and a single lemon; differences in scale, handles and surface finish do not establish objective errors.Confidence: high
Comparison reason · Visual quality: Blind labels: A = Qwen-Image-2.1-Turbo; B = FLUX.2 klein 4B; C = Qwen-Image-2.1; D = Z-Image Turbo. D has convincing ceramic surfaces and wood grain; B is clean with smooth glaze and consistent focus. A and C have slightly synthetic, textured ceramic reflections.Confidence: medium
Comparison reason · Composition: Blind labels: A = Qwen-Image-2.1-Turbo; B = FLUX.2 klein 4B; C = Qwen-Image-2.1; D = Z-Image Turbo. D's diagonal row and clearly foregrounded lemon make the relations easiest to read; B balances the larger teapot well. A is clear but top-heavy with blank wall; C's large cups crowd each other and dominate the pot.Confidence: medium
Comparison reason · Aesthetic appeal: Blind labels: A = Qwen-Image-2.1-Turbo; B = FLUX.2 klein 4B; C = Qwen-Image-2.1; D = Z-Image Turbo. D's natural daylight and B's polished still-life arrangement are equally appealing; A is calm but less integrated, while C's cup scale feels less graceful.Confidence: medium

Cat on the fridge

A black cat sitting on top of a white refrigerator, a brown dog sitting on the floor looking up at it, and exactly four magnets on the fridge door: a star, a heart, a banana and the letter K

Qwen-Image-2.1, Cat on the fridge, seed 11 — *Qwen-Image-2.1*

Z-Image Turbo, Cat on the fridge, seed 11 — *Z-Image Turbo*

FLUX.2 klein 4B, Cat on the fridge, seed 11 — *FLUX.2 klein 4B*Winner

Review reasons

These comparison reasons apply to the set as a whole.

Comparison reason · Prompt adherence: Blind labels: A = Qwen-Image-2.1; B = Qwen-Image-2.1-Turbo; C = FLUX.2 klein 4B; D = Z-Image Turbo. C clearly shows a seated cat, floor-seated brown dog looking steeply toward it, and all four magnets. A and B include everything but their dogs' gaze is shallower. D crops out floor/seated dog evidence and its cat is crouched.Confidence: high
Comparison reason · Correctness: Blind labels: A = Qwen-Image-2.1; B = Qwen-Image-2.1-Turbo; C = FLUX.2 klein 4B; D = Z-Image Turbo. All show four distinct requested magnet shapes, recognizable K letters and coherent visible animal anatomy. A's rotated K remains legible; D's large three-dimensional banana can be a novelty magnet, so its material alone is not a proven error.Confidence: medium
Comparison reason · Visual quality: Blind labels: A = Qwen-Image-2.1; B = Qwen-Image-2.1-Turbo; C = FLUX.2 klein 4B; D = Z-Image Turbo. D gives the most resolved fur and magnet surfaces, C has clean natural detail, B is coherent but the cat is heavily shadowed, and A loses more feline detail to black while adding rough peripheral clutter.Confidence: medium
Comparison reason · Composition: Blind labels: A = Qwen-Image-2.1; B = Qwen-Image-2.1-Turbo; C = FLUX.2 klein 4B; D = Z-Image Turbo. C best connects both animals through the gaze and keeps all requested content framed. B's vertical magnet row is legible; D is a strong close-up but cuts off the requested seated dog. A has more distracting right-edge clutter.Confidence: medium
Comparison reason · Aesthetic appeal: Blind labels: A = Qwen-Image-2.1; B = Qwen-Image-2.1-Turbo; C = FLUX.2 klein 4B; D = Z-Image Turbo. C's bright domestic scene and D's expressive close portrait are comparably appealing. B is calm but stark; A's dim cat and untidy margins are less cohesive.Confidence: medium

Three people

A woman in a green coat holding a red umbrella in her left hand and a brown paper bag in her right hand, standing between a man in a blue suit on her left and a child in a yellow raincoat on her right, in front of a bakery

Qwen-Image-2.1, Three people, seed 11 — *Qwen-Image-2.1*

Z-Image Turbo, Three people, seed 11 — *Z-Image Turbo*

FLUX.2 klein 4B, Three people, seed 11 — *FLUX.2 klein 4B*

Review reasons

These comparison reasons apply to the set as a whole.

Comparison reason · Prompt adherence: Blind labels: A = Qwen-Image-2.1-Turbo; B = FLUX.2 klein 4B; C = Z-Image Turbo; D = Qwen-Image-2.1. Using the woman's anatomical sides, A correctly holds the umbrella in her left and bag in her right, but reverses the man/child positions. B and C reverse both hand assignments and the adjacent people. D has umbrella and bag on her right side while the other hand is pocketed.Confidence: high
Comparison reason · Correctness: Blind labels: A = Qwen-Image-2.1-Turbo; B = FLUX.2 klein 4B; C = Z-Image Turbo; D = Qwen-Image-2.1. A's grips and umbrella shaft are coherent and hand allocation is correct. B and C have wrong hand allocation but coherent visible grips. D's bag hangs from the umbrella-side arm rather than a separate right-hand grip, and the umbrella connection is partly obscured.Confidence: high
Comparison reason · Visual quality: Blind labels: A = Qwen-Image-2.1-Turbo; B = FLUX.2 klein 4B; C = Z-Image Turbo; D = Qwen-Image-2.1. A and C have crisp faces, cloth and hand detail. D's fabrics are strong but the glove/handle region is less clear. B's distant faces and invented storefront lettering are softer and less resolved.Confidence: medium
Comparison reason · Composition: Blind labels: A = Qwen-Image-2.1-Turbo; B = FLUX.2 klein 4B; C = Z-Image Turbo; D = Qwen-Image-2.1. B gives the clearest full trio and bakery context with balanced spacing. A provides readable faces but crops the bag. C crops the canopy and gives the man's back much prominence. D hides both companion faces and the bag grip.Confidence: medium
Comparison reason · Aesthetic appeal: Blind labels: A = Qwen-Image-2.1-Turbo; B = FLUX.2 klein 4B; C = Z-Image Turbo; D = Qwen-Image-2.1. A's restrained portrait and C's lively interaction are equally cohesive. B's friendly tableau and D's moody street scene are appealing but less visually resolved or open.Confidence: medium

5 of 7

Transparency

WinnerQwen-Image-2.1-Turbo. Qwen-Image-2.1-Turbo narrowly led Qwen-Image-2.1; this was a close result. Only the two Qwen models produced substantial genuine transparency in these sets. Opaque white or painted checkerboard backgrounds were not counted as transparency. Glass and smoke still showed compositing defects.

Sticker

A die-cut sticker of a grinning orange cat wearing sunglasses, thick white border, bold vector cartoon style

Qwen-Image-2.1, Sticker, seed 11 — *Qwen-Image-2.1*

Z-Image Turbo, Sticker, seed 11 — *Z-Image Turbo*

FLUX.2 klein 4B, Sticker, seed 11 — *FLUX.2 klein 4B*

Review reasons

These comparison reasons apply to the set as a whole.

Comparison reason · Prompt adherence: Blind labels: A = Qwen-Image-2.1-Turbo; B = FLUX.2 klein 4B; C = Z-Image Turbo; D = Qwen-Image-2.1. A and D most clearly deliver the broad grin and flat bold vector treatment; B and C have cheerful open mouths and slightly more rendered sticker shading. All include the cat, sunglasses and white cut border.Confidence: medium
Comparison reason · Correctness: Blind labels: A = Qwen-Image-2.1-Turbo; B = FLUX.2 klein 4B; C = Z-Image Turbo; D = Qwen-Image-2.1. All have coherent stylized cat faces, connected sunglasses and readable mouths; differences in cartoon teeth and head-only versus bust framing are not objective errors.Confidence: high
Comparison reason · Visual quality: Blind labels: A = Qwen-Image-2.1-Turbo; B = FLUX.2 klein 4B; C = Z-Image Turbo; D = Qwen-Image-2.1. A combines clean vector contours with real alpha (58.62%). D also has real alpha (55.91%) but less consistent outline/border finishing. C is smooth but opaque; B additionally bakes a checkerboard into its opaque output.Confidence: high
Comparison reason · Composition: Blind labels: A = Qwen-Image-2.1-Turbo; B = FLUX.2 klein 4B; C = Z-Image Turbo; D = Qwen-Image-2.1. A and C give the face a strong centered hierarchy with comfortable margins. D remains clear but its uneven upper border pulls attention; B devotes more area to the plain torso and background.Confidence: medium
Comparison reason · Aesthetic appeal: Blind labels: A = Qwen-Image-2.1-Turbo; B = FLUX.2 klein 4B; C = Z-Image Turbo; D = Qwen-Image-2.1. A's broad grin and restrained flat shapes are especially cohesive; C is polished and friendly. D has characterful asymmetry but less refined lines; B's shading and drop shadow are less unified with the flat cartoon.Confidence: medium

Sprite

A full-body game character sprite of a goblin merchant carrying an oversized backpack full of potions, hand-painted style, facing right

Qwen-Image-2.1, Sprite, seed 11 — *Qwen-Image-2.1*

Z-Image Turbo, Sprite, seed 11 — *Z-Image Turbo*

FLUX.2 klein 4B, Sprite, seed 11 — *FLUX.2 klein 4B*

Review reasons

These comparison reasons apply to the set as a whole.

Comparison reason · Prompt adherence: Blind labels: A = FLUX.2 klein 4B; B = Qwen-Image-2.1; C = Z-Image Turbo; D = Qwen-Image-2.1-Turbo. D and B most clearly use painterly modeling with oversized potion-filled packs and rightward full-body poses. D's readable upright merchant is especially complete. A is also close but has an extra visible foot; C is more flat graphic than hand-painted.Confidence: medium
Comparison reason · Correctness: Blind labels: A = FLUX.2 klein 4B; B = Qwen-Image-2.1; C = Z-Image Turbo; D = Qwen-Image-2.1-Turbo. D has the clearest connected limbs and pack straps. C's stylized anatomy is coherent. B's trailing cloth ends in a boot-like form that makes the feet ambiguous. A visibly shows three feet beneath the two-legged torso.Confidence: high
Comparison reason · Visual quality: Blind labels: A = FLUX.2 klein 4B; B = Qwen-Image-2.1; C = Z-Image Turbo; D = Qwen-Image-2.1-Turbo. D and B have usable real alpha (73.67% and 71.69%); D's silhouette and extremities are cleaner. C's graphic rendering is clean but opaque. A is also opaque and its extra foot is a consequential rendering defect.Confidence: medium
Comparison reason · Composition: Blind labels: A = FLUX.2 klein 4B; B = Qwen-Image-2.1; C = Z-Image Turbo; D = Qwen-Image-2.1-Turbo. D separates face, fist, pack and boots clearly. C's simplified colors read well at sprite scale; A's tall stance is legible but the feet crowd together. B's deep crouch and congested cargo compress the hierarchy.Confidence: medium
Comparison reason · Aesthetic appeal: Blind labels: A = FLUX.2 klein 4B; B = Qwen-Image-2.1; C = Z-Image Turbo; D = Qwen-Image-2.1-Turbo. D's warm leather, green skin and jewel-colored bottles form a cohesive painted character. A has rich illustrative personality despite its foot error. C is charming but flatter; B's congested silhouette is less appealing.Confidence: low

Glass

A wine glass half full of red wine with a thin wisp of smoke rising beside it, studio lighting

Qwen-Image-2.1, Glass, seed 11 — *Qwen-Image-2.1*

Z-Image Turbo, Glass, seed 11 — *Z-Image Turbo*

FLUX.2 klein 4B, Glass, seed 11 — *FLUX.2 klein 4B*Winner

Review reasons

These comparison reasons apply to the set as a whole.

Comparison reason · Prompt adherence: Blind labels: A = FLUX.2 klein 4B; B = Z-Image Turbo; C = Qwen-Image-2.1-Turbo; D = Qwen-Image-2.1. A best combines roughly half-full red wine, studio highlights and a thin wisp beside the glass. D places smoke beside it but the plume is broad and the wine is below half the bowl. B puts smoke inside/above the wine. C's beside-glass smoke is scarcely visible in the alpha composite.Confidence: medium
Comparison reason · Correctness: Blind labels: A = FLUX.2 klein 4B; B = Z-Image Turbo; C = Qwen-Image-2.1-Turbo; D = Qwen-Image-2.1. A has coherent glass geometry, a plausible fill level and the requested smoke relationship. C has coherent glass and fill but smoke visibility is weak. D is distinctly underfilled and its plume is broad. B's smoke visibly originates in the wine rather than beside the glass.Confidence: medium
Comparison reason · Visual quality: Blind labels: A = FLUX.2 klein 4B; B = Z-Image Turbo; C = Qwen-Image-2.1-Turbo; D = Qwen-Image-2.1. C and D supply real alpha (91.18% and 91.05%); C transmits the background through the empty bowl more convincingly, while D retains a hazy fill there. Both smoke masks are imperfect. A has cleaner photographic surfaces but no alpha; B is also opaque and embeds a checkerboard.Confidence: medium
Comparison reason · Composition: Blind labels: A = FLUX.2 klein 4B; B = Z-Image Turbo; C = Qwen-Image-2.1-Turbo; D = Qwen-Image-2.1. A balances the full glass against an offset smoke curl. B's centered column remains readable, though the smoke competes with the rim. C leaves substantial blank upper space because its smoke fades away; D shrinks the glass beneath an oversized plume.Confidence: medium
Comparison reason · Aesthetic appeal: Blind labels: A = FLUX.2 klein 4B; B = Z-Image Turbo; C = Qwen-Image-2.1-Turbo; D = Qwen-Image-2.1. A's restrained highlights and delicate offset smoke are most elegant. B is polished but its dark smoke dominates. C's red bowl is attractive yet the intended smoke becomes visually absent; D feels unbalanced and hazy.Confidence: medium

6 of 7

High resolution

WinnerQwen-Image-2.1-Turbo. Qwen-Image-2.1-Turbo led both magazine-cover comparisons, followed by Qwen-Image-2.1. The original-resolution review still found small-text and barcode artifacts. This round has just one prompt and two seeds, so it gives limited evidence about high-resolution performance.

Magazine cover

A fashion magazine cover. Masthead at the top reads "VERDE". A model in an emerald silk suit stands in a field of tall grass at golden hour. Cover lines on the left read "The slow fashion issue" and "48 pages of field notes". A small barcode sits in the bottom right corner with the text "October 2026 · $9".

Qwen-Image-2.1, Magazine cover, seed 11 — *Qwen-Image-2.1*

Z-Image Turbo, Magazine cover, seed 11 — *Z-Image Turbo*

FLUX.2 klein 4B, Magazine cover, seed 11 — *FLUX.2 klein 4B*

Review reasons

These comparison reasons apply to the set as a whole.

Comparison reason · Prompt adherence: Blind labels: A = Qwen-Image-2.1-Turbo; B = FLUX.2 klein 4B; C = Z-Image Turbo; D = Qwen-Image-2.1. A supplies the requested masthead, both exact left cover lines, emerald suit, golden field and exact October 2026 · $9 caption; its sheen reads more clearly as silk than D's relatively matte jacket. D otherwise meets the same content and layout. B has the right scene and satin tailoring but changes the headline and garbles the date caption. C has the main lines but a wrong date/price caption, a more turquoise suit and an unrequested fragmented figure/lettering across the bottom.Confidence: medium
Comparison reason · Correctness: Blind labels: A = Qwen-Image-2.1-Turbo; B = FLUX.2 klein 4B; C = Z-Image Turbo; D = Qwen-Image-2.1. A and D accurately render VERDE, The slow fashion issue, 48 pages of field notes and October 2026 · $9, with coherent visible figures. Their incidental barcode digits contain odd glyphs, but no particular barcode number was specified. B visibly reads fashioun and insture, and its barcode date is malformed. C retains the two headline phrases but replaces the date/price with gibberish beginning Edethuer 20 and has a detached extra torso/hand-like fragment at the bottom. That major spatial failure puts C last.Confidence: high
Comparison reason · Visual quality: Blind labels: A = Qwen-Image-2.1-Turbo; B = FLUX.2 klein 4B; C = Z-Image Turbo; D = Qwen-Image-2.1. A and D have comparably clean typography, detailed hair, continuous suit construction and consistent warm lighting. B has convincing glossy fabric and detailed grasses, but visibly damaged letterforms and small barcode text reduce finish. C has extensive lower-edge compositing debris, stray lettering and fragmented clothing, outweighing its reasonably rendered upper portrait.Confidence: high
Comparison reason · Composition: Blind labels: A = Qwen-Image-2.1-Turbo; B = FLUX.2 klein 4B; C = Z-Image Turbo; D = Qwen-Image-2.1. A and D balance the oversized masthead, clear left text column and dominant figure against quiet sky; both integrate the barcode cleanly at bottom right. Neither has a meaningful overall layout advantage. B keeps a strong figure but places large text over busy grass, reducing legibility and separating the two cover lines widely. C's bold left typography is readable, yet bottom fragments create an unintended competing focal area and undermine cover framing.Confidence: medium
Comparison reason · Aesthetic appeal: Blind labels: A = Qwen-Image-2.1-Turbo; B = FLUX.2 klein 4B; C = Z-Image Turbo; D = Qwen-Image-2.1. A and D share a cohesive restrained green-and-gold editorial treatment; A's direct gaze and D's turned pose offer different but equally persuasive cover moods. B's richer shine and luminous field appeal, though congested text and malformed words make the cover less refined. C's heavy typography and cooler suit are less harmonious here, and conspicuous bottom artifacts break the image.Confidence: medium

7 of 7

Anatomy

WinnerQwen-Image-2.1. Qwen-Image-2.1 narrowly led Qwen-Image-2.1-Turbo. No Eight-Angle result clearly met every pose requirement; lotus images also showed obscured limb connections or apparent extra feet. A higher rank here means fewer visible errors, not anatomically correct output.

Eight-Angle Pose

A realistic full-body sports photograph of an adult woman performing Eight-Angle Pose (Astavakrasana) on a yoga mat in a bright plain studio. She wears an opaque fitted teal athletic top and black leggings. Both palms are planted shoulder-width apart on the mat, fingers spread, elbows bent near 90 degrees, supporting her entire body. Her torso leans forward almost horizontal. Both legs extend together toward her right side, almost parallel to the floor, with her right upper arm threaded between her thighs and her ankles crossed. Her hips and both feet are lifted clear of the floor. Show her entire body, both hands and both feet inside the frame, from a three-quarter front view. Natural human anatomy, clear joint connections, realistic balance, soft daylight, no mirrors, no other people, no text.

Qwen-Image-2.1, Eight-Angle Pose, seed 11 — *Qwen-Image-2.1*

Z-Image Turbo, Eight-Angle Pose, seed 11 — *Z-Image Turbo*

FLUX.2 klein 4B, Eight-Angle Pose, seed 11 — *FLUX.2 klein 4B*

Review reasons

These comparison reasons apply to the set as a whole.

Comparison reason · Prompt adherence: Blind labels: A = Qwen-Image-2.1; B = Qwen-Image-2.1-Turbo; C = Z-Image Turbo; D = FLUX.2 klein 4B. B keeps both extended feet lifted and overlapping; A also lifts both but separates the ankles. Neither bends elbows near 90 degrees or threads the right upper arm between thighs. D is a grounded plank; C additionally bends one leg upward and drops the torso steeply.Confidence: high
Comparison reason · Correctness: Blind labels: A = Qwen-Image-2.1; B = Qwen-Image-2.1-Turbo; C = Z-Image Turbo; D = FLUX.2 klein 4B. D and C show traceable limbs and visible supporting foot contact, though neither is the requested pose. A and B place almost all torso/leg mass behind nearly vertical supporting arms, giving an unsupported floating appearance; their required arm-thigh relationship is absent. No defensible anatomical separation within either pair.Confidence: medium
Comparison reason · Visual quality: Blind labels: A = Qwen-Image-2.1; B = Qwen-Image-2.1-Turbo; C = Z-Image Turbo; D = FLUX.2 klein 4B. A and B render cloth, hair, skin and hand edges crisply with controlled studio light. D is similarly photographic but has softer overlapping foot detail. C has lower fine-detail clarity and less even surface rendering.Confidence: medium
Comparison reason · Composition: Blind labels: A = Qwen-Image-2.1; B = Qwen-Image-2.1-Turbo; C = Z-Image Turbo; D = FLUX.2 klein 4B. B frames the entire silhouette and mat with useful breathing room. D provides a clear three-quarter face and hand arrangement. A hides more of the face and compresses the body into an arch; C's overlapping bent leg and downward head make the intended joint relationships hardest to read.Confidence: medium
Comparison reason · Aesthetic appeal: Blind labels: A = Qwen-Image-2.1; B = Qwen-Image-2.1-Turbo; C = Z-Image Turbo; D = FLUX.2 klein 4B. B's restrained neutral light and D's clean bright sports-photo treatment are comparably appealing. A is more cramped in its body silhouette; C's steep arch and busy limb crossing feel less graceful.Confidence: low

Full Lotus Pose

A realistic full-body sports photograph of an adult woman meditating in full Lotus Pose (Padmasana) on a yoga mat in a bright plain studio. She wears an opaque fitted teal athletic top and black leggings. She sits upright with a straight spine, relaxed shoulders and closed eyes. Both knees rest low toward the mat. Her legs are folded and crossed in full lotus: her right foot rests high on her left thigh and her left foot rests high on her right thigh, both soles facing upward. Show both feet clearly, each connected to its own lower leg, and the natural overlapping arrangement of both crossed legs. Her hands rest gently palm-up on her knees, each thumb touching the index finger and the other three fingers relaxed. Show her entire body including both hands and both feet, from a slightly elevated three-quarter front view. Natural human anatomy, clear joint connections, realistic balance, soft daylight, no mirrors, no other people, no text.

Qwen-Image-2.1, Full Lotus Pose, seed 11 — *Qwen-Image-2.1*Winner

Z-Image Turbo, Full Lotus Pose, seed 11 — *Z-Image Turbo*

FLUX.2 klein 4B, Full Lotus Pose, seed 11 — *FLUX.2 klein 4B*

Review reasons

These comparison reasons apply to the set as a whole.

Comparison reason · Prompt adherence: Blind labels: A = FLUX.2 klein 4B; B = Qwen-Image-2.1-Turbo; C = Qwen-Image-2.1; D = Z-Image Turbo. C alone shows both feet on top of opposite thighs and a three-quarter view, but its open hands do not make the requested thumb-index circles. B clearly supplies those circles but leaves one foot tucked underneath. A also keeps a foot low in front with raised knees; D substitutes a simple crossed-leg seat with both feet low.Confidence: high
Comparison reason · Correctness: Blind labels: A = FLUX.2 klein 4B; B = Qwen-Image-2.1-Turbo; C = Qwen-Image-2.1; D = Z-Image Turbo. C's two visible feet and crossed-leg paths are broadly consistent with lotus, although a rear ankle segment is occluded and the mudra is wrong. B has coherent half-lotus limbs. D is a coherent ordinary crossed seat. A's foreground foot has compressed elongated toe shapes and its raised-knee foot placement is less clear. None receives credit for hidden connections as if fully visible.Confidence: medium
Comparison reason · Visual quality: Blind labels: A = FLUX.2 klein 4B; B = Qwen-Image-2.1-Turbo; C = Qwen-Image-2.1; D = Z-Image Turbo. B and C have comparably fine cloth, skin and hair detail with clean studio light. A remains sharp overall but its foreground foot and curled fingers are less resolved. D is visibly softer in face, hair and garment detail.Confidence: medium
Comparison reason · Composition: Blind labels: A = FLUX.2 klein 4B; B = Qwen-Image-2.1-Turbo; C = Qwen-Image-2.1; D = Z-Image Turbo. C's three-quarter angle gives the clearest view of both top-side feet and mat depth. B has balanced centering and open hands, but hides a foot. A and D are similarly centered yet their leg overlaps hide the requested placement.Confidence: medium
Comparison reason · Aesthetic appeal: Blind labels: A = FLUX.2 klein 4B; B = Qwen-Image-2.1-Turbo; C = Qwen-Image-2.1; D = Z-Image Turbo. C's gentle diagonal framing and B's quiet symmetry are equally cohesive. A's brighter flatter finish is pleasant but less nuanced; D's softer, flatter capture is less refined.Confidence: low

Contestants

Qwen-Image-2.1-Turbo

Version: bf16 weights from the Comfy-Org repackage, ComfyUI 0.39.0 (commit 0df64eb)
Licence: Qwen Research License Agreement
Files: qwen_image_2.1_turbo_bf16.safetensors, qwen3vl_8b_bf16.safetensors, qwen_image_2.1_vae_bf16.safetensors
guidance: 1
sampler: euler
schedule: the model card's fixed 8-step schedule
steps: 8

Qwen-Image-2.1

Version: bf16 weights from the Comfy-Org repackage, ComfyUI 0.39.0 (commit 0df64eb)
Licence: Qwen Research License Agreement
Files: qwen_image_2.1_bf16.safetensors, qwen3vl_8b_bf16.safetensors, qwen_image_2.1_vae_bf16.safetensors
guidance: 1
sampler: euler
schedule: simple
steps: 40

Z-Image Turbo

Version: bf16 weights from the Comfy-Org repackage, ComfyUI 0.39.0 (commit 0df64eb)
Licence: Apache License 2.0
Files: z_image_turbo_bf16.safetensors, qwen_3_4b.safetensors, ae.safetensors
guidance: 1
sampler: res_multistep
schedule: simple, shift 3
steps: 8

FLUX.2 klein 4B

Version: bf16 weights from the Comfy-Org repackage, ComfyUI 0.39.0 (commit 0df64eb)
Licence: Apache License 2.0
Files: flux-2-klein-4b.safetensors, qwen_3_4b.safetensors, flux2-vae.safetensors
guidance: 1
sampler: euler
schedule: Flux2Scheduler
steps: 4

Changes since publishing

10 October 2026Add per-set aspect rankings, recorded generation time and estimated rental execution cost.

Sources