Muse Wonder scores 70.3 on BuildingBench, ranking #12 of 31 for photo-to-3D
Lianhuiq · x · 2026-10-13
EnactraAI's Muse Wonder (max) scored 70.3 on BuildingBench for photo-to-3D building reconstruction, ranking #12 of 31 — roughly level with GPT-6 Sol. Vs Muse Spark 1.3: score 68.2→70.3, median cost per building $5.89→$7.72, higher geometry/surface but lower appearance scores. The gain is within statistical uncertainty; best reconstruction was Helsinki's Pikkuparlamentti at 81.6. Alexandr Wang said the focus was real-world usefulness over flashy 3D demos.
More from Multimodal
- Kandinsky 6.0 Video goes open source: 3B/29B models generate clips with synced audio — lmoroney · 2026-10-13
- promptpilot MCP server pipes 20+ image, video and audio models into Claude Code — modelcontextprotocol · 2026-10-13
- AI video of a Chinese girl and her giant cat looks 'way too good' — basedjensen · 2026-10-13
- Midjourney v8.2 style code sref 1582397130 delivers painterly water-light photography looks — azed_ai · 2026-10-13
- GPT-6 Astra builds 100% 3D hotel website via Blender MCP with 10.7M triangles — ravisparikh · 2026-10-13
- Midjourney V7 ultraviolet gradient snake eye macro via --sref 1632499533 — michaelrabone · 2026-10-13