Seeking evals for DeepSeek V4.1 Flash quant running on 2x Spark

A user says consensus favors DeepSeek V4.1 Flash over GLM 5.3 Flash Abliterated for the task, but asks whether the quantized build running on 2x Spark is nerfed and requests evals.

Nate Raw

@_naterawArchived 8 days agoOriginally posted 2026.10.24

Nate Raw @_naterawOriginally posted 2026.10.24

Open original on X

Consensus seems to be new DeepSeek v4.1 Flash is enough for this (and is preferred over GLM 5.3 Flash Ablit ckpt)…is the quant that runs DS on 2xSparks enough or is it nerfed? If anyone has evals please share!

This post quotes another post: x.com/thezachmueller/status/2113563988781076752

x.com/_nateraw/status/2113719950732824876Archived 2026.10.24 12:07

Open original on X