DeepSeek-V3 Trained With Only 180K GPU-Hours, Slashing MoE Compute Costs
teortaxesTex · x · 2026-08-01
A comparison of DeepSeek's generational training compute reveals massive efficiency gains: V1 required 300K H800 GPU-hours per 1T tokens, while V2 and V3 dropped to 173K and 180K respectively. The author estimates their next model (V4-Flash) might need only around 66K hours, totaling roughly 2M GPU-hours.
This implies DeepSeek's current best model is smaller and cheaper to train than its version from 26 months ago. The author praises this as the ultimate validation of Scaling laws for MoEs, contrasting DeepSeek's transparent inference unit economics with the secretive myth-making of Western frontier AI companies.
Related event: DeepSeek Drastically Reduces Training Compute Costs Across Models(2 posts)→
More from Infra
- 5TB of Data Stored on a Tiny Glass Slab Marks Microscopic Storage Breakthrough — TansuYegen · 2026-08-01
- antirez Enables Lossless MXFP4 Local Inference for DeepSeek v4 Flash — antirez · 2026-08-01
- Running Qwen 27B on 3x 2080Ti at 55tps: Optimal Config Shared — AccountGotLocked69 · 2026-08-01
- Tata in Talks with ASML to Manufacture Advanced Chipmaking Subassemblies in India — prasanna_says · 2026-08-01
- 3bit 27B Model Runs at Usable Speed for Reinforcement Learning — cephaloform · 2026-08-01
- Stanford's Mark Horowitz Questions the Future of High-Speed Links and Scaling Trends — jwt0625 · 2026-08-01