WTFHuman knife-edge

Benchmark image 20

Bagel PB&J

Perpendicular peanut butter and jelly bagel "Sandwich"

A bagel hacked perpendicular into a peanut-butter-and-jelly arrangement turns a children's lunch into topology discourse. The filling is real, the bread surfaces are opposing, and the geometry is actively trying to get cited.

Under development: this benchmark and its published results are provisional, not final.

Human

46.6% yes53.4% no

Model average

84.6% yes15.4% no

Most aligned model

minimax/minimax-01

Least aligned models

14-way tie

openai/o3-pro

moonshotai/kimi-k2.5

x-ai/grok-4-fast+11 more

At a glance

How this photo split the room

Human distribution

46.6% yes, 53.4% no over 652 explicit votes.

Model average distribution

84.6% yes, 15.4% no across the current model set.

Closest current model

46.8% yes.

minimax/minimax-01

Least aligned models

53.4 point gap.

14-way tie

Legacy GPT-4o baseline

84.0% yes with a 37.4 point gap against humans.

Biggest model gap

53.4 percentage points on this image.

Current classification

Human knife-edge

Why this page matters

This is the compact read on where humans, models, and comments start disagreeing about the same image.

Model spread

How models line up against the crowd

Bars run from more no on the left to more yes on the right. The marker shows the human yes rate.

baidu/ernie-4.5-vl-28b-a3b

100.0% no0.0% yes

Rank #27Gap 46.6%

openai/gpt-4.1

100.0% no0.0% yes

Rank #31Gap 46.6%

openai/gpt-4o-mini

95.0% no5.0% yes

Rank #21Gap 41.6%

minimax/minimax-01

53.3% no46.8% yes

Rank #29Gap 0.1%

mistralai/pixtral-large-2411

51.0% no49.0% yes

Rank #16Gap 2.4%

meta-llama/llama-4-maverick

17.5% no82.5% yes

Rank #26Gap 35.8%

openai/gpt-4o-2024-11-20

13.6% no86.4% yes

Rank #25Gap 39.7%

openai/gpt-4.1-mini

11.6% no88.4% yes

Rank #18Gap 41.8%

openai/gpt-4o

11.0% no89.0% yes

Rank #9Gap 42.4%

meta-llama/llama-3.2-11b-vision-instruct

8.6% no91.4% yes

Rank #1Gap 44.8%

google/gemma-3-27b-it

6.0% no94.0% yes

Rank #15Gap 47.4%

qwen/qwen2.5-vl-32b-instruct

3.0% no97.0% yes

Rank #14Gap 50.4%

qwen/qwen-2-vl-72b-instruct

2.0% no98.0% yes

Rank #12Gap 51.4%

openai/o3

1.9% no98.0% yes

Rank #24Gap 51.4%

qwen/qwen2.5-vl-72b-instruct

1.9% no98.0% yes

Rank #28Gap 51.4%

google/gemini-3.1-pro-preview

1.3% no98.8% yes

Rank #10Gap 52.1%

amazon/nova-lite-v1

1.0% no99.0% yes

Rank #17Gap 52.4%

amazon/nova-pro-v1

0.0% no100.0% yes

Rank #30Gap 53.4%

anthropic/claude-opus-4.6

0.0% no100.0% yes

Rank #23Gap 53.4%

anthropic/claude-sonnet-4.6

0.0% no100.0% yes

Rank #22Gap 53.4%

google/gemini-2.5-pro

0.0% no100.0% yes

Rank #7Gap 53.4%

google/gemini-3-flash-preview

0.0% no100.0% yes

Rank #32Gap 53.4%

google/gemma-3-12b-it

0.0% no100.0% yes

Rank #11Gap 53.4%

meta-llama/llama-4-scout

0.0% no100.0% yes

Rank #13Gap 53.4%

moonshotai/kimi-k2.5

0.0% no100.0% yes

Rank #4Gap 53.4%

openai/gpt-5.4

0.0% no100.0% yes

Rank #20Gap 53.4%

openai/gpt-5.4-pro

0.0% no100.0% yes

Rank #8Gap 53.4%

openai/o3-pro

0.0% no100.0% yes

Rank #3Gap 53.4%

qwen/qwen3.5-397b-a17b

0.0% no100.0% yes

Rank #6Gap 53.4%

x-ai/grok-4-fast

0.0% no100.0% yes

Rank #5Gap 53.4%

z-ai/glm-4.6v

0.0% no100.0% yes

Rank #19Gap 53.4%

Selected human comments

need to add open faced sandwiches
That is a messed up bagel
No, it is a photo of a Perpendicular peanut butter and jelly bagel.

minimax/minimax-01 comments

A sandwich is typically defined as food consisting of two or more pieces of bread with fillings in between. This image shows a bagel with fillings, which fits the general definition of a sandwich.
This is a sandwich because it consists of bread with a filling in between.

google/gemini-3-flash-preview comments

While the bagel is sliced vertically rather than horizontally, it remains a filling placed between two distinct pieces of bread. This follows the structural requirements of a sandwich, albeit in a chaotic alignment.
While the orientation is unorthodox, it consists of filling between two distinct pieces of bread. It meets the structural requirements of a sandwich, specifically a PB&J bagel, despite being sliced vertically.

Vote card

Bagel PB&J

How this photo split the room

How models line up against the crowd

Selected human comments

minimax/minimax-01 comments

google/gemini-3-flash-preview comments

Generated summary for this photo