Evalytic Image Models Leaderboard

Run on fal.ai · 2026-03-12 · 2 models · 1 prompts · Judge: gemini-2.5-flash

| Family: |

VLM Total 80%

Visual Quality 33%

Prompt Adherence 33%

Text Rendering ⓘ 33%

Deterministic Total 20%

CLIP Score 40%

Sharpness 30%

Aesthetic 30%

Image model leaderboard ranked by overall score
#	Model	Family	$/img	Overall	VQ	PA	TR	CLIP	Sharp	Aesthetic	Score/$	imgsys	Time	Seed	fal.ai
1	Z-Image Turbo	Other	$0.005	4.90	5.0	5.0	5.0	0.358	0.943	—	980	1080	2.2s	42	Try →
2	FLUX.1 [schnell]	Black Forest Labs	$0.003	4.55	4.0	5.0	5.0	0.313	0.945	—	1517	950	1.4s	42	Try →

How We Score

2 models evaluated on fal.ai infrastructure
1 curated prompts from PartiPrompts, DrawBench, OneIG-Bench + custom
Judge: gemini-2.5-flash
3 VLM dimensions: Visual Quality, Prompt Adherence, Text Rendering (text prompts only)
Deterministic metrics: CLIP Score, Sharpness, Aesthetic Score
Seed: 2 models with seed=42 (reproducible), 0 models use median of 3 runs (metadata-derived from model roster)
Resolution: landscape_4_3 (all models), fal.ai default parameters
Default weighting: 80% VLM + 20% deterministic (editorial default, customizable above)
Overall formula: 1 + 4 × (vlm_w × (vlm_avg − 1)/4 + det_w × det_avg)
Date: 2026-03-12

Link copied!