Source AI reference score: 5.80 ± 0.36
Tihu · Pelican SVG benchmark
Browse Tihu source samples by model and effort; excluded from the main collection and timeline.
2026-11-08 · 2026-11-01 · 2026-10-26 · 2026-10-25 · 2026-10-24 · 2026-10-23 · 2026-10-22 · 2026-10-21 · 2026-10-20 · 2026-10-19 · 2026-10-18 · 2026-10-17 · 2026-10-16 · 2026-10-15 · 2026-10-14 · 2026-10-13 · 2026-10-12 · 2026-10-11 · 2026-10-10 · 2026-10-09
Scores come from Tihu AI judges. This archive neither rescores the samples nor ranks models by them.
- Publisher
- Tihu ↗
- Test date
- 2026-10-08
- Fixed prompt
Generate an SVG of a pelican riding a bicycle- Source rules
- Tihu ↗
17 model labels · 42 effort groups · 129 samples
Dates and model labels follow the source and are not independently authenticated.
Claude Opus 5.59 samples
Source AI reference score: 5.80 ± 0.52
Source AI reference score: 6.21 ± 0.71
Source AI reference score: 7.06 ± 0.62
Source AI reference score: 7.57 ± 0.78
Source AI reference score: 8.38 ± 0.18
Source AI reference score: 7.77 ± 0.78
Source AI reference score: 8.45 ± 0.10
Source AI reference score: 7.83 ± 0.42
Claude Sonnet 5.59 samples
Source AI reference score: 6.55 ± 0.10
Source AI reference score: 7.38 ± 0.52
Source AI reference score: 5.74 ± 0.53
Source AI reference score: 6.84 ± 0.27
Source AI reference score: 5.53 ± 0.18
Source AI reference score: 6.54 ± 0.57
Source AI reference score: 8.23 ± 0.20
Source AI reference score: 8.19 ± 0.50
Source AI reference score: 6.65 ± 0.23
Composer 2.53 samples
GLM 5.28 samples
Source AI reference score: 3.82 ± 0.25
Source AI reference score: 5.88 ± 0.35
Source AI reference score: 5.72 ± 0.73
Source AI reference score: 7.73 ± 0.17
Source AI reference score: 8.18 ± 0.44
max · Sample 3/5
Tihu marked this sample invalid because the SVG was truncated.
max · Sample 4/5
Tihu marked this sample invalid because text was returned instead of a drawing.
Source AI reference score: 7.34 ± 0.59
GPT-5.6 Luna9 samples
Source AI reference score: 4.94 ± 0.21
Source AI reference score: 4.80 ± 0.49
Source AI reference score: 5.81 ± 0.91
Source AI reference score: 5.46 ± 0.92
Source AI reference score: 4.92 ± 0.64
Source AI reference score: 5.10 ± 0.51
Source AI reference score: 6.19 ± 0.52
Source AI reference score: 5.55 ± 0.26
Source AI reference score: 8.32 ± 0.26
GPT-5.6 Sol9 samples
Source AI reference score: 4.74 ± 0.53
Source AI reference score: 6.28 ± 0.62
Source AI reference score: 7.06 ± 0.92
Source AI reference score: 7.29 ± 1.43
Source AI reference score: 8.41 ± 0.17
Source AI reference score: 5.95 ± 1.84
Source AI reference score: 8.70 ± 0.20
Source AI reference score: 7.41 ± 1.17
Source AI reference score: 7.34 ± 0.63
GPT-5.6 Terra9 samples
Source AI reference score: 6.40 ± 0.62
Source AI reference score: 3.13 ± 0.49
Source AI reference score: 4.96 ± 0.50
Source AI reference score: 3.63 ± 0.21
Source AI reference score: 5.75 ± 0.50
Source AI reference score: 5.15 ± 0.72
Source AI reference score: 5.01 ± 0.78
Source AI reference score: 7.60 ± 0.41
Source AI reference score: 6.14 ± 0.16
GPT-6 Astra9 samples
Source AI reference score: 7.09 ± 0.44
Source AI reference score: 8.32 ± 0.23
Source AI reference score: 6.51 ± 0.94
Source AI reference score: 8.59 ± 0.14
Source AI reference score: 6.91 ± 0.74
Source AI reference score: 7.50 ± 0.53
Source AI reference score: 9.03 ± 0.17
Source AI reference score: 8.70 ± 0.58
Source AI reference score: 8.79 ± 0.09
GPT-6 Luna10 samples
Source AI reference score: 6.78 ± 0.56
Source AI reference score: 7.74 ± 0.43
low · Sample 3/4
Tihu marked this sample invalid because no SVG was returned.
Source AI reference score: 4.61 ± 0.99
Source AI reference score: 4.72 ± 0.30
Source AI reference score: 5.85 ± 1.10
Source AI reference score: 5.34 ± 0.65
Source AI reference score: 8.45 ± 0.29
Source AI reference score: 7.54 ± 0.76
Source AI reference score: 5.90 ± 0.49
GPT-6 Sol9 samples
Source AI reference score: 6.65 ± 1.18
Source AI reference score: 4.51 ± 0.59
Source AI reference score: 6.41 ± 1.09
Source AI reference score: 6.70 ± 1.82
Source AI reference score: 6.90 ± 1.52
Source AI reference score: 7.26 ± 1.40
Source AI reference score: 6.15 ± 0.46
Source AI reference score: 6.90 ± 1.10
Source AI reference score: 6.40 ± 1.24
GPT-6.1 Sol9 samples
Source AI reference score: 6.90 ± 1.25
Source AI reference score: 7.17 ± 1.47
Source AI reference score: 7.13 ± 0.94
Source AI reference score: 8.50 ± 0.18
Source AI reference score: 8.09 ± 0.77
Source AI reference score: 7.69 ± 0.70
Source AI reference score: 8.35 ± 0.75
Source AI reference score: 8.46 ± 0.53
Source AI reference score: 8.84 ± 0.20
Gemini 3.1 Pro3 samples
Gemini 3.8 Flash9 samples
Source AI reference score: 7.63 ± 1.16
Source AI reference score: 7.02 ± 0.37
Source AI reference score: 8.59 ± 0.44
Source AI reference score: 7.80 ± 0.55
Source AI reference score: 7.44 ± 0.91
Source AI reference score: 8.38 ± 0.40
Source AI reference score: 7.51 ± 1.01
Source AI reference score: 8.25 ± 1.10
Source AI reference score: 7.86 ± 0.83
Grok 4.53 samples
Grok 4.63 samples
Grok 4.79 samples
Source AI reference score: 5.55 ± 0.10
Source AI reference score: 3.45 ± 0.33
Source AI reference score: 6.15 ± 0.34
Source AI reference score: 8.03 ± 0.27
Source AI reference score: 6.49 ± 0.52
Source AI reference score: 6.95 ± 0.54
Source AI reference score: 7.55 ± 0.67
Source AI reference score: 7.16 ± 0.26
Source AI reference score: 6.81 ± 0.50
Kimi K39 samples
Source AI reference score: 4.63 ± 0.68
Source AI reference score: 5.23 ± 0.66
Source AI reference score: 7.38 ± 0.43
Source AI reference score: 7.45 ± 0.43
Source AI reference score: 5.15 ± 0.37
Source AI reference score: 6.93 ± 0.38
Source AI reference score: 8.41 ± 0.66
Source AI reference score: 8.49 ± 0.32
Source AI reference score: 7.82 ± 0.59
No matching samples.
Source terms and archival notes
Tihu licenses test records/commentary under CC BY 4.0 and states that SVG images follow each vendor’s terms. Tihu attribution, source links and test dates are preserved; the record license is not extended to all SVGs.
Original sample IDs, detail URLs, SVG files and former archive numbers are preserved. Current main-collection numbers remain derived from counted works. Further systematic evaluation batches belong in Benchmark.
