Character Recognition on IlluChar Semantic
98.7Accuracy (Large)Human
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Human2026.03 | 98.7 | 99.3 | 100 | 99 | — | |
| Gemini-2.5-Pro2026.03 | 4.7 | 4.6 | 12 | 5.2 | -89.7 | |
| GPT-5.22026.03 | 4 | 1.4 | 4 | 3.3 | -85.8 | |
| GLM-4.5V2026.03 | 3.6 | 6 | 13.3 | 4.9 | -90.4 | |
| Claude-Sonnet-4.52026.03 | 3 | 3.2 | 9.3 | 3.5 | -84 | |
| Qwen3-VL-235B-A22B2026.03 | 2.7 | 7.8 | 6.7 | 4.4 | -93.3 | |
| Qwen3-VL-8B2026.03 | 2.4 | 6.4 | 6.7 | 3.8 | -89.4 |