source
Scale Labs
Benchmarks · /benchmarks/scale-vista
VISTA
Vision-language evaluation on the Scale leaderboard.
Passport
Thin verified coverageThis is a rubric-judged signal, so it is more structured than arena taste but still depends on the scoring rubric.
metric
Score (%)
judge
Rubric
direction
higher better
group id
scale_vista_current
domain
Vision understanding
What it measures vs what it misses
✓ Measures
Visual reasoning and understanding.
✗ Misses
Image generation or editing quality.
Why this countsIt is useful when the model must read charts, UI, screenshots, or visual scenes rather than text alone.Comparable-group ruleThis percentile only compares models inside the exact benchmark/version group shown here. It is not a universal score.What it missesIt does not tell you whether the model can generate or edit images well.
Leaderboard · this benchmark version
#1 · Gemini 2.5 Pro
SL · Apr 1, 2026
82%
#2 · GPT-5
SL · Apr 1, 2026
79%
#3 · GPT-5.4
SL · Apr 1, 2026
79%
#4 · GPT-5.4 mini
SL · Apr 1, 2026
79%
#5 · GPT-5.4 nano
SL · Apr 1, 2026
79%
#6 · Claude Opus 4.1
SL · Apr 1, 2026
77%
#7 · Claude Opus 4
SL · Apr 1, 2026
77%
#8 · Claude Opus 4.6
SL · Apr 1, 2026
77%
#9 · Claude Opus 4.7
SL · Apr 1, 2026
77%
#10 · Claude Sonnet 4
SL · Apr 1, 2026
75%
#11 · Claude Sonnet 4.5
SL · Apr 1, 2026
75%
#12 · Claude Sonnet 4.6
SL · Apr 1, 2026
75%
#13 · Grok 4
SL · Apr 1, 2026
73%
#14 · Qwen3 235B A22B
SL · Apr 1, 2026
71%
#15 · Llama 4 Maverick
SL · Apr 1, 2026
69%