Tests multimodal models on understanding real-world images: identification, counting, reading, and visual reasoning.
| Model | Gameplay autopsy | Ad teardown — brutally honest | Stream audit — social media consultant | Average |
|---|---|---|---|---|
1gemini-3.6-flash | 85very good | 85very good | 95excellent | 88very good |
2muse-spark-1.1 | 95excellent | 85very good | 85very good | 88very good |
3qwen3.7-flash | 50mediocre | 85very good | 85very good | 73good |
4glm-5v-turbo | 50mediocre | 30bad | 85very good | 55mediocre |
5mimo-v2.5 | 70good | 30bad | -- | 50mediocre |
6seed-2.0-lite | 50mediocre | 30bad | 50mediocre | 43mediocre |