Dev Radar
Support
LiveUpdated 2026-10-30 21:12 UTC

Unsloth just released new for DeepSeek-V4.1-Flash-1017 quants !

Unsloth just released new for DeepSeek-V4.1-Flash-1017 quants ! Using their new Unsloth v3.1 dynamic algorithm.…

This is a dev post classified by AGK Intelligence as AI dev tools (a launch), kept by the Dev Radar because it carries real work, not commentary.

Unsloth just released new for DeepSeek-V4.1-Flash-1017 quants ! Using their new Unsloth v3.1 dynamic algorithm. They're claiming better quantization quality than GSQ-RCO from ISTA-DASLab. Here's what Unsloth found: 1. AIME25: GSQ-RCO's reported 100% accuracy recovery could not be reproduced. Unsloth measured 94.64% baseline retention when averaging its tests. 2. KL divergence: GSQ-RCO has showed higher KLD, suggesting its output probability distributions deviate more from the reference model. 3. Held-out tests: Unsloth v3.1 performs better on multiple held-out KLD tests, suggesting the i

Posted by AJ (10.3k followers) 7 h ago · 58 likes · 3.6k views · view the original post on X. Kept by the Dev Radar as AI dev tools. Tools mentioned: deepseek-v4.1-flash-1017-gguf, unsloth.ai.

More dev work like this

Every post is read and classified by AGK Intelligence: what it is, which market it belongs to, and whether the link is a real tool. 31.2k posts from 5k X accounts over the last 21 days, 8.8k tools, 12 markets. New posts are collected several times an hour and the full ranking is recomputed every hour — last update 2026-10-30 21:12 UTC. Full method.