DeepSeek V4.1 Flash enters BuildingBench top five at 1/15 of Sonnet's cost
Lianhuiq · x · 2026-10-18
EnactraAI tested DeepSeek V4.1 Flash 1017 on 3D building reconstruction from real photos via BuildingBench, calling it an exciting leap for open models.
- Overall score of 81.9, up 9 points from the previous version's 72.9, landing in the top five
- Slightly above Fable 5.1's 81.4 and essentially level with Sonnet 5.5's 82.1
- Estimated median cost of $3.01/task — about 1/11 of Fable 5.1's and 1/15 of Sonnet 5.5's cost
More from Models
- "$400/Month Can't Buy $40k of API Forever": Team Benchmarks Open Models as Fallback — ChrisRackauckas · 2026-10-18
- OpenAI's model list surfaces aurora-red/green/blue-alpha checkpoints, tipped as Fable 5.5 answer — ChrisGPT · 2026-10-18
- Solo dev's after-hours decision model tops open-weights on JevBench, beating commercial rivals — lukaszkaiser · 2026-10-18
- DeepSeek V4.1 Flash jumps on long-run agentic tasks but regresses on constraints — teortaxesTex · 2026-10-18
- Open models outperform on routine code maintenance, but only Opus finds deep defects — ChrisRackauckas · 2026-10-18
- MOSS SFT vs Flux 3 Klein vs SD 3.0: Same-Prompt Comparison Exposes SD3's Age — ZootAllures9111 · 2026-10-18