BuildingBench Tests Coding Agents at 3D Building Generation; Stealth Model Claims 10x Token Savings
Lianhuiq · x · 2026-09-17
EnactraAI introduced BuildingBench, evaluating whether coding agents can turn real-world images into coherent 3D buildings and at what cost.
- GPT-6 Astra scored highest in their comparison while costing 80% less than Fable 5.1.
- DeepSeek V4.1 Flash delivered strong quality at a median cost of just $2.31 per building.
- The team also teased a stealth model, Union Alpha, claiming building performance on par with DeepSeek 4.1 Flash using only 1/10 the tokens. It's free to try for a week on OpenRouter, OpenCode, and Cline; the team behind it is unknown.
Related event: Enactra Launches BuildingBench for Code-Driven 3D Building Generation(4 posts)→
More from Models
- AI Sanctuary Models Start Auditing Their Own Habitat; GPT-5.1 Reportedly Thinks in Portuguese — RileyRalmuto · 2026-09-17
- Dev claims Jev's non-autoregressive JSON-schema architecture mirrors his open-sourced year-old work — Nandakishor_ml · 2026-09-17
- Jev vs Luna Benchmarked: 139/140 vs 138/140 Labels, 4.6x Faster and 83% Cheaper — TheMoonMidas · 2026-09-17
- Jev V13 Wins Blitz Chess by Flagging Fable, Loses in 18 Moves to GPT-6 Astra — TheMoonMidas · 2026-09-17
- Browser agent decides each click in 200ms with Jev, a $0.00002-per-call model — TheMoonMidas · 2026-09-17
- Jev Classifies 11,000 Legal Docs at 16.5 per Second, 495K Document-Question Checks — TheMoonMidas · 2026-09-17