vLLM Silently Corrupted DeepSeek-V4.1 Engram Weights via 8-bit Quantization

tokenbender · x · 2026-11-19

Alibaba engineers discovered a nasty quantization bug while running DeepSeek-V4.1 on T-Head Zhenwu 810E chips.

tokenbender draws the broader lesson: NN training is insanely forgiving of catastrophic-scale bugs — this could have hidden inside a 4-day cluster RL run and you'd never know why performance was bad.

Original post →

More from Infra

Infra channel →