Spatial reasoning on SIBench single-image
58.86AccuracyGPT-5
Evaluation Results
| Method | Links | |
|---|---|---|
| GPT-5Model Type=Understanding-only MLLM2025.12 | 58.86 | |
| Qwen3VL-32BModel Type=Understanding-only MLLM2025.12 | 55.49 | |
| InternVL3.5-38BModel Type=Understanding-only MLLM2025.12 | 53.61 | |
| Qwen3VL-8BModel Type=Understanding-only MLLM2025.12 | 50.13 | |
| InternVL3.5-8BModel Type=Understanding-only MLLM2025.12 | 50.13 | |
| COOPERModel Type=Unified MLLM2025.12 | 50.07 | |
| GPT-4oModel Type=Understanding-only MLLM2025.12 | 49.48 | |
| BAGEL-PEModel Type=Unified MLLM, Variant=Perception Enhancement2025.12 | 48.25 | |
| BAGEL-REModel Type=Unified MLLM, Variant=Reasoning Enhancement2025.12 | 43.98 | |
| BAGELModel Type=Unified MLLM2025.12 | 43.57 | |
| Janus-Pro-7BModel Type=Unified MLLM2025.12 | 42.73 | |
| Liquid-7BModel Type=Unified MLLM2025.12 | 38.63 |