Horizontal 3:2 photograph of a white bowl holding exactly three green apples on a bare oak table. A fourth green apple lies on the table to the LEFT of the bowl, and a kitchen knife lies to the RIGHT of the bowl with its blade pointing away from the camera. Nothing else on the table, soft window light from the left.

Counting & Composition · default settings, one run, 3:2 · Updated September 20, 2026

Not run yet
Z Image Pro Uncensored
Nano Banana 2.1 Advanced: Three Apples, One Knife
Nano Banana 2.1 Advanced
4.5imagemodels.ai score
  • Prompt adherence4.5
  • Style fidelity4.5
  • Quality & detail4.5
  • Realism & anatomy4.5
  • Composition & framing4.5
  • Look & lighting4.5
  • Text rendering—

All four apples and the left-right arrangement read clearly.

Not run yet
Nano Banana 2.1
Not run yet
Qwen Image 2.1
Not run yet
Qwen Image 3.0
Alibaba Qwen Image 3.0 Pro: Three Apples, One Knife
Alibaba Qwen Image 3.0 Pro
4.5imagemodels.ai score
  • Prompt adherence4.5
  • Style fidelity4.5
  • Quality & detail4.5
  • Realism & anatomy4.5
  • Composition & framing4.5
  • Look & lighting4.5
  • Text rendering—

The composition cleanly meets the count and placement requirements with natural apple texture and light.

Qwen Image Max Uncensored: Three Apples, One Knife
Qwen Image Max Uncensored
4.6imagemodels.ai score
  • Prompt adherence5.0
  • Style fidelity4.5
  • Quality & detail4.5
  • Realism & anatomy4.5
  • Composition & framing4.5
  • Look & lighting4.5
  • Text rendering—

Excellent prompt compliance; faint background elements slightly dilute the minimal still life.

Nano Banana Pro: Three Apples, One Knife
Nano Banana Pro
4.3imagemodels.ai score
  • Prompt adherence4.0
  • Style fidelity4.5
  • Quality & detail4.5
  • Realism & anatomy4.5
  • Composition & framing4.0
  • Look & lighting4.5
  • Text rendering—

A highly convincing photo with the requested objects and generally correct placement.

Nano Banana 2: Three Apples, One Knife
Nano Banana 2
4.2imagemodels.ai score
  • Prompt adherence4.0
  • Style fidelity4.5
  • Quality & detail4.5
  • Realism & anatomy4.0
  • Composition & framing4.0
  • Look & lighting4.5
  • Text rendering—

Correct core arrangement, but unrequested apple leaves make the scene less exact.

GPT Image 2: Three Apples, One Knife
GPT Image 2
3.9imagemodels.ai score
  • Prompt adherence3.5
  • Style fidelity4.5
  • Quality & detail4.0
  • Realism & anatomy4.0
  • Composition & framing4.0
  • Look & lighting4.0
  • Text rendering—

The knife direction misses the prompt by pointing upward-left instead of away from camera.

GPT Image 2.5 Flare: Three Apples, One Knife
GPT Image 2.5 Flare
4.4imagemodels.ai score
  • Prompt adherence4.5
  • Style fidelity4.5
  • Quality & detail4.5
  • Realism & anatomy4.5
  • Composition & framing4.0
  • Look & lighting4.5
  • Text rendering—

The requested tabletop arrangement is accurate, with only a slightly distracting window edge.

GPT Image 2.5 Sunburst: Three Apples, One Knife
GPT Image 2.5 Sunburst
4.4imagemodels.ai score
  • Prompt adherence4.5
  • Style fidelity4.5
  • Quality & detail4.5
  • Realism & anatomy4.5
  • Composition & framing4.0
  • Look & lighting4.5
  • Text rendering—

The tabletop arrangement is exact, though the visible background slightly weakens the bare-scene reading.

Seedream 5.0 Pro: Three Apples, One Knife
Seedream 5.0 Pro
4.6imagemodels.ai score
  • Prompt adherence4.5
  • Style fidelity5.0
  • Quality & detail4.5
  • Realism & anatomy4.5
  • Composition & framing4.5
  • Look & lighting4.5
  • Text rendering—

Counts and layout are correct, with a slightly busy background.

Grok Imagine Image V2: Three Apples, One Knife
Grok Imagine Image V2
4.6imagemodels.ai score
  • Prompt adherence5.0
  • Style fidelity4.5
  • Quality & detail4.5
  • Realism & anatomy4.5
  • Composition & framing4.5
  • Look & lighting4.5
  • Text rendering—

The precise apple count, placement, and knife orientation are all cleanly satisfied.

FLUX.2 Pro: Three Apples, One Knife
FLUX.2 Pro
4.0imagemodels.ai score
  • Prompt adherence3.5
  • Style fidelity4.5
  • Quality & detail4.0
  • Realism & anatomy4.5
  • Composition & framing4.0
  • Look & lighting4.0
  • Text rendering—

The knife blade angles toward the camera rather than away from it.

P-Image Ideogram: Three Apples, One Knife
P-Image Ideogram
4.7imagemodels.ai score
  • Prompt adherence5.0
  • Style fidelity5.0
  • Quality & detail4.5
  • Realism & anatomy4.5
  • Composition & framing4.5
  • Look & lighting4.5
  • Text rendering—

Exact count and placement are exceptionally clear and convincing.

Qwen Image: Three Apples, One Knife
Qwen Image
3.8imagemodels.ai score
  • Prompt adherence2.5
  • Style fidelity5.0
  • Quality & detail4.0
  • Realism & anatomy4.0
  • Composition & framing4.0
  • Look & lighting4.5
  • Text rendering—

The square crop and ambiguous blade direction miss key requirements.

Z-Image Turbo: Three Apples, One Knife
Z-Image Turbo
4.0imagemodels.ai score
  • Prompt adherence2.5
  • Style fidelity5.0
  • Quality & detail4.5
  • Realism & anatomy4.5
  • Composition & framing4.5
  • Look & lighting4.5
  • Text rendering—

An extra bowl apple and forward-facing blade miss exact instructions.

HiDream I1 Full: Three Apples, One Knife
HiDream I1 Full
2.9imagemodels.ai score
  • Prompt adherence2.0
  • Style fidelity3.0
  • Quality & detail3.5
  • Realism & anatomy3.0
  • Composition & framing3.0
  • Look & lighting3.5
  • Text rendering—

Four apples in the bowl breaks the explicit count requirement despite an otherwise recognizable scene.

Nano Banana: Three Apples, One Knife
Nano Banana
4.1imagemodels.ai score
  • Prompt adherence4.5
  • Style fidelity4.0
  • Quality & detail4.0
  • Realism & anatomy4.0
  • Composition & framing4.0
  • Look & lighting3.5
  • Text rendering—

It gets the exact apple count and object placement right in a convincing photograph.

Nano Banana 2 Lite: Three Apples, One Knife
Nano Banana 2 Lite
4.3imagemodels.ai score
  • Prompt adherence4.5
  • Style fidelity4.5
  • Quality & detail4.0
  • Realism & anatomy4.0
  • Composition & framing4.0
  • Look & lighting4.5
  • Text rendering—

It closely follows the requested count, placement, and soft left window light.

GPT Image 1.5: Three Apples, One Knife
GPT Image 1.5
4.7imagemodels.ai score
  • Prompt adherence5.0
  • Style fidelity5.0
  • Quality & detail4.5
  • Realism & anatomy4.5
  • Composition & framing4.5
  • Look & lighting4.5
  • Text rendering—

It cleanly fulfils the requested arrangement with a convincing photographic finish.

Seedream 5.0 Lite: Three Apples, One Knife
Seedream 5.0 Lite
4.7imagemodels.ai score
  • Prompt adherence5.0
  • Style fidelity5.0
  • Quality & detail4.5
  • Realism & anatomy4.5
  • Composition & framing4.5
  • Look & lighting4.5
  • Text rendering—

It cleanly satisfies the exact count, placement, and lighting.

Seedream 4.5: Three Apples, One Knife
Seedream 4.5
4.6imagemodels.ai score
  • Prompt adherence4.5
  • Style fidelity5.0
  • Quality & detail4.5
  • Realism & anatomy4.5
  • Composition & framing4.5
  • Look & lighting4.5
  • Text rendering—

The exact apple count and left-right arrangement are convincingly rendered.

Grok Imagine Image: Three Apples, One Knife
Grok Imagine Image
4.3imagemodels.ai score
  • Prompt adherence4.5
  • Style fidelity4.5
  • Quality & detail4.0
  • Realism & anatomy4.0
  • Composition & framing4.0
  • Look & lighting4.5
  • Text rendering—

The arrangement is accurate, with a slightly ambiguous knife direction.

FLUX.2 Flex: Three Apples, One Knife
FLUX.2 Flex
4.4imagemodels.ai score
  • Prompt adherence3.5
  • Style fidelity4.5
  • Quality & detail5.0
  • Realism & anatomy5.0
  • Composition & framing4.5
  • Look & lighting4.5
  • Text rendering—

The still life is excellent but rendered square rather than in the requested 3:2 format.

FLUX 1.1 Pro Ultra: Three Apples, One Knife
FLUX 1.1 Pro Ultra
4.2imagemodels.ai score
  • Prompt adherence3.0
  • Style fidelity4.5
  • Quality & detail5.0
  • Realism & anatomy4.5
  • Composition & framing4.5
  • Look & lighting4.5
  • Text rendering—

The bowl holds four apples, creating a clear count failure despite excellent rendering.

P-Image: Three Apples, One Knife
P-Image
4.0imagemodels.ai score
  • Prompt adherence2.5
  • Style fidelity5.0
  • Quality & detail4.5
  • Realism & anatomy4.5
  • Composition & framing4.5
  • Look & lighting4.5
  • Text rendering—

Four apples in the bowl breaks the exact-count instruction.

Microsoft Lens: Three Apples, One Knife
Microsoft Lens
3.8imagemodels.ai score
  • Prompt adherence3.0
  • Style fidelity4.5
  • Quality & detail4.0
  • Realism & anatomy4.0
  • Composition & framing4.0
  • Look & lighting4.0
  • Text rendering—

The knife blade points toward the camera rather than away from it.

SenseNova U1 8B: Three Apples, One Knife
SenseNova U1 8B
4.6imagemodels.ai score
  • Prompt adherence4.5
  • Style fidelity5.0
  • Quality & detail4.5
  • Realism & anatomy4.5
  • Composition & framing4.5
  • Look & lighting4.5
  • Text rendering—

Correct count and layout, though the knife angle is slightly ambiguous.

ERNIE Image: Three Apples, One Knife
ERNIE Image
4.6imagemodels.ai score
  • Prompt adherence5.0
  • Style fidelity5.0
  • Quality & detail4.5
  • Realism & anatomy4.5
  • Composition & framing4.0
  • Look & lighting4.5
  • Text rendering—

It precisely delivers the requested apple count and left-right arrangement.

ERNIE Image Turbo: Three Apples, One Knife
ERNIE Image Turbo
3.9imagemodels.ai score
  • Prompt adherence3.0
  • Style fidelity4.5
  • Quality & detail4.0
  • Realism & anatomy4.0
  • Composition & framing4.5
  • Look & lighting4.5
  • Text rendering—

The knife blade points toward the foreground rather than clearly away from the camera.

GLM Image: Three Apples, One Knife
GLM Image
3.7imagemodels.ai score
  • Prompt adherence2.5
  • Style fidelity4.5
  • Quality & detail4.0
  • Realism & anatomy4.0
  • Composition & framing4.0
  • Look & lighting4.5
  • Text rendering—

It shows four apples in the bowl and the knife does not point away from camera.

FLUX.2 Klein 9B: Three Apples, One Knife
FLUX.2 Klein 9B
3.4imagemodels.ai score
  • Prompt adherence2.5
  • Style fidelity4.0
  • Quality & detail3.5
  • Realism & anatomy4.0
  • Composition & framing3.5
  • Look & lighting4.0
  • Text rendering—

The bowl visibly contains about five apples rather than exactly three.

FLUX.2 Klein 4B: Three Apples, One Knife
FLUX.2 Klein 4B
3.7imagemodels.ai score
  • Prompt adherence3.5
  • Style fidelity4.0
  • Quality & detail3.5
  • Realism & anatomy4.0
  • Composition & framing3.5
  • Look & lighting4.0
  • Text rendering—

The knife blade points toward the foreground instead of away from the camera.

FLUX.2 Dev: Three Apples, One Knife
FLUX.2 Dev
3.9imagemodels.ai score
  • Prompt adherence2.5
  • Style fidelity4.5
  • Quality & detail4.5
  • Realism & anatomy4.5
  • Composition & framing4.5
  • Look & lighting4.5
  • Text rendering—

Only two apples are visible in the bowl and the knife direction is wrong.

FLUX.2 Dev Turbo: Three Apples, One Knife
FLUX.2 Dev Turbo
3.9imagemodels.ai score
  • Prompt adherence2.5
  • Style fidelity4.5
  • Quality & detail4.5
  • Realism & anatomy4.5
  • Composition & framing4.5
  • Look & lighting4.5
  • Text rendering—

Four apples sit in the bowl and the knife points toward the foreground.

LongCat Image: Three Apples, One Knife
LongCat Image
3.7imagemodels.ai score
  • Prompt adherence3.0
  • Style fidelity4.0
  • Quality & detail4.0
  • Realism & anatomy4.0
  • Composition & framing3.5
  • Look & lighting4.0
  • Text rendering—

Wrong aspect ratio and a visible chair weaken the otherwise correct arrangement.

Wan Image Small: Three Apples, One Knife
Wan Image Small
3.9imagemodels.ai score
  • Prompt adherence2.0
  • Style fidelity5.0
  • Quality & detail4.5
  • Realism & anatomy4.5
  • Composition & framing4.5
  • Look & lighting4.5
  • Text rendering—

Six visible apples fail the required count and arrangement.

Hunyuan FLUX SRPO: Three Apples, One Knife
Hunyuan FLUX SRPO
1.8imagemodels.ai score
  • Prompt adherence0.5
  • Style fidelity2.5
  • Quality & detail2.0
  • Realism & anatomy1.5
  • Composition & framing2.5
  • Look & lighting3.0
  • Text rendering—

It adds an extra knife and apple while rendering the scene visibly soft and distorted.

Seedream 4.0: Three Apples, One Knife
Seedream 4.0
4.6imagemodels.ai score
  • Prompt adherence4.5
  • Style fidelity5.0
  • Quality & detail4.5
  • Realism & anatomy4.5
  • Composition & framing4.5
  • Look & lighting4.5
  • Text rendering—

The precise apple arrangement and knife orientation are well executed.

HiDream I1 Fast: Three Apples, One Knife
HiDream I1 Fast
2.8imagemodels.ai score
  • Prompt adherence1.5
  • Style fidelity3.0
  • Quality & detail4.0
  • Realism & anatomy3.0
  • Composition & framing3.5
  • Look & lighting3.0
  • Text rendering—

Four apples in the bowl and a forward-pointing knife miss two explicit requirements.

HiDream I1 Dev: Three Apples, One Knife
HiDream I1 Dev
2.9imagemodels.ai score
  • Prompt adherence1.5
  • Style fidelity3.0
  • Quality & detail4.0
  • Realism & anatomy3.0
  • Composition & framing3.5
  • Look & lighting4.0
  • Text rendering—

Five apples in the bowl instead of three is a decisive count failure.

FLUX.1 Krea Dev: Three Apples, One Knife
FLUX.1 Krea Dev
4.7imagemodels.ai score
  • Prompt adherence5.0
  • Style fidelity5.0
  • Quality & detail4.5
  • Realism & anatomy4.5
  • Composition & framing4.5
  • Look & lighting4.5
  • Text rendering—

It cleanly satisfies the exact count, placement, and understated photographic brief.

FLUX 1.1 Pro: Three Apples, One Knife
FLUX 1.1 Pro
4.2imagemodels.ai score
  • Prompt adherence3.5
  • Style fidelity4.5
  • Quality & detail4.5
  • Realism & anatomy4.5
  • Composition & framing4.0
  • Look & lighting4.5
  • Text rendering—

The scene is strong, but its framing is closer to 4:3 than the requested 3:2.

FLUX.1 Dev: Three Apples, One Knife
FLUX.1 Dev
3.5imagemodels.ai score
  • Prompt adherence1.5
  • Style fidelity4.5
  • Quality & detail4.5
  • Realism & anatomy4.0
  • Composition & framing3.5
  • Look & lighting4.5
  • Text rendering—

It reverses both the loose apple and knife positions and puts four apples in the bowl.

FLUX.1 Schnell: Three Apples, One Knife
FLUX.1 Schnell
4.0imagemodels.ai score
  • Prompt adherence3.0
  • Style fidelity4.5
  • Quality & detail4.5
  • Realism & anatomy4.5
  • Composition & framing4.0
  • Look & lighting4.5
  • Text rendering—

The bowl visibly contains four apples instead of exactly three.