deepseek-v4.1POST
Unsloth Q4_K_XL: DeepSeek V4.1 Flash needs 500.7GB vs Qwen's 111.3GB
A quantization size comparison notes Unsloth's Q4_K_XL build of V4.1 Flash 1017 takes 500.7GB versus 111.3GB for the Qwen model.
Steven | SelfHostedSI
@SelfHostedSIArchived 10 days agoOriginally posted 2026.10.22
Steven | SelfHostedSI @SelfHostedSIOriginally posted 2026.10.22
@TheAhmadOsman Also the exact reverse of what they need in memory: Unsloth's Q4_K_XL is 500.7GB for V4.1 Flash 1017 and 111.3GB for the Qwen.
x.com/selfhostedsi/status/2113073257498997026Archived 2026.10.22 11:59