deepseek-v4.1POST
Commenter pushes back: V4.1 Flash 1017 is not GPT 5.6 Sol, quantized even less so
In a thread, a commenter agrees quantized models cover most needs but insists DeepSeek V4.1 Flash 1017 still trails 5.6 Sol, and the quantized version is further off.
Rezha Adrian Tanuharja
@JustATypicGuyArchived 12 days agoOriginally posted 2026.10.19
Rezha Adrian Tanuharja @JustATypicGuyOriginally posted 2026.10.19
@GeoWashingtn @jpschroeder "quant models can do the vast majority of what people actually need." yeah but Justin's point remains: DS V4.1 Flash 1017 is not like 5.6 Sol, and the quantized version in your reply is even farther from Sol
x.com/justatypicguy/status/2112025264322740239Archived 2026.10.19 12:02