AnalysisAI ModelsNovember 2, 2026

Unsloth and ISTA's GSQ-RCO clash over DeepSeek-V4.1 GGUF benchmarks

Read original source →reddit.com

Unsloth says its DeepSeek-V4.1 Dynamic v3.1 GGUFs beat ISTA's GSQ-RCO quants by >10-20% on DeepSWE, Terminal Bench, Automation Bench, HLE and SciCode at 512K context. A Reddit user then flagged that all four Unsloth quants plus the Q8 baseline rose ~2-3 points on LiveCodeBench v6 at once.

How this story unfolded

4 days · 0 reports · 4 community posts · from Oct 30

  1. Oct 30
  2. Nov 2
  3. Nov 3

More stories today

Open the live feed