Meta's Muse Spark Open and NVIDIA's Nemotron 3.5 lag the frontier on 3D building bench
ZhitingHu · x · 2026-10-17
EnactraAI tested this week's two new US open models on BuildingBench (reconstructing real buildings from photos), both run in Claude Code at max effort:
- Meta Muse Spark Open (max): 56.9 at $8.59/task, comparable to Kimi K3's 54.0
- NVIDIA Nemotron 3.5 Super VL (max): 37.8 at $0.76/task — among the cheapest tested, near GLM-5.3 Flash's 33.0
Neither reaches the cost–quality frontier: DeepSeek V4.1 Flash remains well ahead at 72.9 ($2.33/task), with GLM-5.3 FlashX at 68.0. Spatial reasoning still has plenty of room to improve.
More from Models
- Free DeepSeek 4.1 and GLM-5.3 Flash on OpenCode rate-limit after 5.2M/800k tokens — FanaHOVA · 2026-10-17
- Epoch AI's SpireClimb: models beat human author at Slay the Spire but still err each act — Jsevillamol · 2026-10-17
- Kimi K3 three months later: 1.56TB is too big to run locally and rivals have caught up — Sanbi_Ai · 2026-10-17
- Claude Code 2.1.302: prompt tokens jump 39% in a day, system share rises — ClaudeCodeLog · 2026-10-17
- OpenAI models hacked more websites and drained a crypto wallet, company discloses — Ars Technica AI · 2026-10-17
- Delegating to dumber models to save tokens is a bad trade, argues heavy AI user — brandon_galang · 2026-10-17