⚡ AI Visual Benchmark

PokéBench

The visual coding benchmark for LLMs. Every model renders a complete trading card in pure SVG. AI-judged. Ranked.

19Benchmarks
19Models
1Themes
glm-5 low effort
87s · 5.7K tok · Mar 01, 2026 · ★ 91
grok-4.1-fast low effort
36s · 4.4K tok · Mar 01, 2026 · ★ 85
gpt-oss-120b low effort
34s · 2.6K tok · Mar 01, 2026 · ★ 74
gpt-5.2 low effort
88s · 7.7K tok · Mar 01, 2026 · ★ 92
gpt-5-nano low effort
20s · 4.0K tok · Mar 01, 2026 · ★ 85
kimi-k2.5 low effort
194s · 9.5K tok · Mar 01, 2026 · ★ 90
minimax-m2.5 low effort
47s · 4.4K tok · Mar 01, 2026 · ★ 90
minimax-m2.1 low effort
102s · 6.4K tok · Mar 01, 2026 · ★ 83
gemini-3.1-pro-preview low effort
46s · 3.8K tok · Mar 01, 2026 · ★ 87
gemini-3-flash-preview low effort
20s · 4.2K tok · Mar 01, 2026 · ★ 60
gemini-2.5-flash low effort
39s · 11.0K tok · Mar 01, 2026 · ★ 87
gemini-2.5-flash-lite low effort
25s · 9.3K tok · Mar 01, 2026 · ★ 80
gemini-2.0-flash-lite-001 low effort
17s · 2.8K tok · Mar 01, 2026 · ★ 84
deepseek-v3.2 low effort
98s · 3.8K tok · Mar 01, 2026 · ★ 86
claude-sonnet-4-6✦ Max low effort
100s · 8.7K tok · Mar 01, 2026 · ★ 92
claude-sonnet-4-5✦ Max low effort
100s · 8.6K tok · Mar 01, 2026 · ★ 90
claude-opus-4-6✦ Max low effort
71s · 5.7K tok · Mar 01, 2026 · ★ 89
trinity-large-preview:free low effort
137s · 2.9K tok · Mar 01, 2026 · ★ 81
claude-haiku-4.5 low effort
48s · 9.5K tok · Mar 01, 2026 · ★ 87