Web Navigation on DeepShop (live)
65.8AccuracyWeblica-8B
Evaluation results
| Method | Links | |
|---|---|---|
| Weblica-8BTotal # Steps=240, k=8, Model Type=Open-weight2026.05 | 65.8 | |
| Gemini CUATotal # Steps=100, Model Type=API-only2026.05 | 62 | |
| Weblica-8BTotal # Steps=120, k=4, Model Type=Open-weight2026.05 | 55.9 | |
| Weblica-8BTotal # Steps=60, k=2, Model Type=Open-weight2026.05 | 45.4 | |
| MolmoWeb-8BTotal # Steps=≥100†, Model Type=Open-weight2026.05 | 42.3 | |
| Weblica-8BTotal # Steps=30, k=1, Model Type=Open-weight2026.05 | 34.2 | |
| GLM-4.1V-9B-ThinkingTotal # Steps=100, Model Type=Open-weight2026.05 | 32 | |
| Fara-7BTotal # Steps=≥100†, Model Type=Open-weight2026.05 | 26.2 | |
| OpenAI CUATotal # Steps=100, Model Type=API-only2026.05 | 24.7 | |
| Qwen3-VL-Instruct-8BTotal # Steps=30, Model Type=Open-weight2026.05 | 24.1 | |
| UI-TARS-1.5-7BTotal # Steps=100, Model Type=Open-weight2026.05 | 11.6 |