GUI Web Agent Navigation on Mind2Web Online
67.3Overall Average ScoreOpenWebRL-4B w/ OpenWebRL-Judge-8B
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| OpenWebRL-4B w/ OpenWebRL-Judge-8Bbackbone=4B, # Steps=30, # Tasks=2.2k2026.06 | 67.3 | — | — | — | — | — | — | — | — | |
| OpenWebRL-4Bbackbone=4B, # Steps=30, # Tasks=2.2k2026.06 | 67 | — | — | — | — | — | — | — | — | |
| OpenWebRL-8Bbackbone=8B, # Steps=30, # Tasks=2.2k2026.06 | 67 | — | — | — | — | — | — | — | — | |
| Qwen3-VL-235B-A22B-Thinking# Steps=30, # Tasks=–2026.06 | 63.7 | — | — | — | — | — | — | — | — | |
| OpenAI computer-use-preview# Steps=100, # Tasks=–2026.06 | 58.3 | — | — | — | — | — | — | — | — | |
| GPT-5 (SoM)# Steps=100, # Tasks=–2026.06 | 57.7 | — | — | — | — | — | — | — | — | |
| Gemini computer-use-preview# Steps=100, # Tasks=–2026.06 | 57.3 | — | — | — | — | — | — | — | — | |
| o3 (SoM)# Steps=100, # Tasks=–2026.06 | 55.4 | — | — | — | — | — | — | — | — | |
| OpenWebRL-8B-SFTbackbone=8B, # Steps=30, # Tasks=0.9k2026.06 | 54 | — | — | — | — | — | — | — | — | |
| OpenWebRL-4B-SFTbackbone=4B, # Steps=30, # Tasks=0.4k2026.06 | 47 | — | — | — | — | — | — | — | — | |
| Gemini-3-flash (Axtree)# Steps=100, # Tasks=–2026.06 | 44.8 | — | — | — | — | — | — | — | — | |
| GPT-5 (Axtree)# Steps=30, # Tasks=–2026.06 | 41.9 | — | — | — | — | — | — | — | — | |
| Qwen3-VL-8B-Thinking# Steps=30, # Tasks=–2026.06 | 38.7 | — | — | — | — | — | — | — | — | |
| MolmoWeb-8B# Steps=100, # Tasks=>278.5k2026.06 | 35.3 | — | — | — | — | — | — | — | — | |
| Gemini-3-flash (Axtree)# Steps=30, # Tasks=–2026.06 | 34.8 | — | — | — | — | — | — | — | — | |
| GPT-4o (SoM)# Steps=100, # Tasks=–2026.06 | 34.6 | — | — | — | — | — | — | — | — | |
| Fara-7B# Steps=100, # Tasks=>123.2k2026.06 | 34.1 | — | — | — | — | — | — | — | — | |
| GLM-4.1V-9B-Thinking# Steps=100, # Tasks=–2026.06 | 33.9 | — | — | — | — | — | — | — | — | |
| Qwen3-VL-4B-Thinking# Steps=30, # Tasks=–2026.06 | 32 | — | — | — | — | — | — | — | — | |
| UI-TARS-1.5-7B# Steps=100, # Tasks=–2026.06 | 31.3 | — | — | — | — | — | — | — | — | |
| MolmoWeb-4B# Steps=100, # Tasks=>278.5k2026.06 | 31.3 | — | — | — | — | — | — | — | — | |
| GPT-4.1 + UGround-v1-7BSteps=302026.02 | 29.7 | 41.3 | 21.7 | 5.2 | 22.7 | 47.5 | 35 | 28.6 | 36.7 | |
| GPT-5 + UGround-v1-7BSteps=302026.02 | 29.7 | 40 | 24.5 | 14.3 | 26 | 45 | 31.5 | 24.7 | 33.3 | |
| GUI-Libra-8BSteps=302026.02 | 28 | 31.3 | 17.5 | 10.4 | 19.3 | 42.5 | 37.8 | 28.6 | 36.7 | |
| Qwen3-VL-32BSteps=302026.02 | 26.8 | 33.8 | 17.5 | 7.8 | 19.3 | 45 | 31.5 | 28.6 | 34.3 | |
| GUI-Libra-4BSteps=302026.02 | 25.7 | 36.3 | 18.2 | 6.5 | 20 | 45 | 30.1 | 19.5 | 31.3 | |
| GUI-Libra-7BSteps=302026.02 | 25.5 | 36.3 | 15.4 | 2.6 | 17.7 | 47.5 | 30.8 | 23.4 | 33.3 | |
| GPT-4o + UGround-v1-7BSteps=302026.02 | 24.3 | 38.8 | 14 | 6.5 | 18.7 | 45 | 27.3 | 19.5 | 30 | |
| ScaleCUA-7BSteps=302026.02 | 23.7 | 33.8 | 14.7 | 3.9 | 17 | 47.5 | 27.3 | 18.2 | 30.3 | |
| ScaleCUA-32BSteps=302026.02 | 23.5 | 31.3 | 14.7 | 6.5 | 17 | 43.8 | 29.4 | 16.9 | 30 | |
| Qwen3-VL-4BSteps=302026.02 | 21.7 | 33.8 | 10.5 | 6.5 | 15.7 | 43.8 | 23.8 | 18.2 | 27.7 | |
| GUI-Libra-3BSteps=302026.02 | 21.3 | 28.8 | 9.8 | 5.2 | 13.7 | 47.5 | 21 | 24.7 | 29 | |
| Qwen3-VL-8BSteps=302026.02 | 19.3 | 23.8 | 9.8 | 0 | 11 | 43.8 | 23.1 | 19.5 | 27.7 | |
| ScaleCUA-3BSteps=302026.02 | 16.3 | 30 | 4.9 | 2.6 | 11 | 37.5 | 18.2 | 11.7 | 21.7 | |
| Qwen2.5-VL-7BSteps=302026.02 | 15.8 | 22.5 | 7.7 | 0 | 9.7 | 36.3 | 18.9 | 13 | 22 | |
| Qwen2.5-VL-32BSteps=302026.02 | 13.5 | 12.5 | 7.7 | 1.3 | 7.3 | 28.8 | 14.7 | 19.5 | 19.7 | |
| Qwen2.5-VL-3BSteps=302026.02 | 4.8 | 3.8 | 0 | 1.3 | 1.3 | 16.3 | 7.7 | 1.3 | 8.3 |