Benchmarks

How well each model draws pixel art through the plugin, on the same fixed tasks.

Leaderboard

Each model's best score per benchmark, averaged over the benchmarks it has been run on.

Compliance ranks first; craft, judged 0–4 on five axes, breaks ties. Read the rubric.

Models ranked by their mean best score
#	Model	Score	Craft	Coverage	Runs
1	claude-opus-5-5	98%	35%	3/3 benchmarks	6

All benchmarks

Boombox mage — character and anime sound-blast attack
A street-mage with a magic boombox drawn to a fixed design, then an anime-style special attack where he blasts a sound wave out of the boombox.

validate
palette
read
spec
shading
tagged
layers
anticipation
impact
wave
recoil
dissipate
timing
on model

1 claude-opus-5-5

2 steps, 2 ranked runs
Cherry tree — from seed to full bloom
A blossoming cherry tree drawn to a fixed design, then animated growing from a seed into exactly that tree.

validate
palette
read
spec
shading
tagged
final
stages
continuous
ground
easing
timing

1 claude-opus-5-5

2 steps, 2 ranked runs
Winding road — deep-perspective landscape and ambient loop
A valley landscape with a road winding into the distance, drawn to a fixed layout, then a seamless ambient loop with a wagon travelling the road toward the viewer.

validate
palette
read
spec
road
depth
frames
seamless
static
windmill
river
smoke
wagon

1 claude-opus-5-5

2 steps, 2 ranked runs

1. Fixed prompts

Same words, canvas and palette for every model.
2. Run in Aseprite

The model draws through the plugin, one session per step.
3. Scored

Each criterion is a yes-or-no check on the finished files.

Benchmarks are run by the maintainers; new models are added as they are tested. A prompt's revision changes when its wording or criteria do, and only runs on the current revision are ranked.