Epoch AI - Blueprint Bench 2: leaderboard

Epoch AI Blueprint-Bench 2 evaluates spatial reasoning: agents reconstruct a 2D floor plan of an apartment from ~20 interior photographs, scored by graph-based similarity of room layout and connectivity to the ground truth.

Metric: Score. Source: epoch.ai. 33 models tracked.

Top models

#ModelScoreOverall rank
1gemini-4-argon_unknown0.54
2claude-opus-5-5_unknown0.51
3gpt-6-astra_unknown0.5
4claude-fable-5-1_unknown0.42
5claude-sonnet-5-5_unknown0.39
6claude-fable-5_unknown0.39
7gemini-3.8-flash_unknown0.39
8gpt-6-sol_unknown0.37
9gpt-5.5_unknown0.36
10gemini-3.5-flash_unknown0.34

Interactive version: theaggregate.ai/benchmark?slug=epoch-ai-blueprint-bench-2 · How It Works · Data refreshed daily, snapshot 2026-10-20.