List your company

Spatial Reasoning on MapEval

72.7AccuracyGPT-5
44.151.52558.9566.375Jun 11, 2026

Evaluation results

MethodLinks
GPT-5Model Type=Commercial2026.0672.7
Gemini-2.5-FlashModel Type=Commercial2026.0671
Gemini-2.5-ProModel Type=Commercial2026.0671
Qwen3-235BModel Type=Open-Source2026.0669
Qwen3-32BModel Type=Open-Source2026.0667.2
Qwen3-30BModel Type=Open-Source2026.0664.7
w/o ToolModel Type=Ablation2026.0659
PERIA-8BModel Type=Proposed2026.0655.5
w/o RLModel Type=Ablation2026.0654.8
InternVL3.5-14BModel Type=Open-Source2026.0654
Qwen3-8BModel Type=Backbone Baseline2026.0654
InternVL3.5-8BModel Type=Open-Source2026.0653.7
w/o SFTModel Type=Ablation2026.0649.5
Mini-o3Model Type=Tool-Augmented2026.0646.7
VTool-R1-7BModel Type=Tool-Augmented2026.0645.7
R1-OneVision-7BModel Type=Tool-Augmented2026.0645.2