Game-based visual-agent performance on Super Mario Bros
950.5Total Performance ScoreGLM-5V
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GLM-5VSkill condition=MMSkills2026.05 | 950.5 | 1,384.5 | |
| GLM-5VSkill condition=Text-only2026.05 | 794.5 | 1,218 | |
| Qwen3-VL-235BSkill condition=MMSkills2026.05 | 788 | 1,514.25 | |
| Qwen3-VL-8B-InstructSkill condition=MMSkills2026.05 | 764 | 1,128.75 | |
| Gemini 3 FlashSkill condition=MMSkills2026.05 | 624 | 1,081.33 | |
| GLM-5VSkill condition=No skill2026.05 | 612.75 | 1,191.5 | |
| Qwen3-VL-235BSkill condition=Text-only2026.05 | 610.5 | 1,138.25 | |
| Qwen3-VL-8B-InstructSkill condition=Text-only2026.05 | 596.5 | 997.25 | |
| Gemini 3 FlashSkill condition=Text-only2026.05 | 548 | 912 | |
| Qwen3-VL-235BSkill condition=No skill2026.05 | 454.5 | 955.5 | |
| Qwen3-VL-8B-InstructSkill condition=No skill2026.05 | 415.25 | 928.75 | |
| Gemini 3 FlashSkill condition=No skill2026.05 | 411 | 766.67 |