Qwen 1-bit Quantization Shrinks Model to 397GB, a 91% Reduction

danielhanchen · x · 2026-08-13

UnslothAI has successfully further quantized the Qwen model down to 397GB. Compared to the standard IQ1S size of 508GB, this represents a massive 91% reduction while still maintaining surprisingly good performance. The new quantization is now supported in Unsloth Desktop with Canvas mode.

Related event: Unsloth Shrinks Qwen3.8 by 91% for Local Deployment(3 posts)→

Original post →

More from Models

Models channel →