WHAT NEW IN X10.31 SAT

AI collections from X · 22 collections · 61,986 posts

Back to No.020 DeepSeek V4.1 Flash hands-on tests and comparisonsUpdated 15 hours ago
deepseek-v4.1THREAD

DeepSeek V4.1-Flash-1017 runs 3-bit locally at 100 tok/s in a new HF repo

The author published a 3-bit quantized pack for V4.1-Flash-1017 that runs locally at about 100 tok/s, their first Hugging Face repo.

Nate’s in SF
@NateemersonArchived 12 days agoOriginally posted 2026.10.18
THREADPosted in 2 parts

  1. I built a pack for DeepSeek-v4.1-Flash-1017 Running 3bit at 100 tok/s locally My first huggingface repo!

    This post quotes another post: x.com/deepseek_ai/status/2111375640717033777

    DeepSeek V4.1-Flash-1017 runs 3-bit locally at 100 tok/s in a new HF repo (thread part 1)

  2. Working on some AB tests with real benches now. Please let me know if you see anything I did wrong or should fix, hoping to do more of this in the future! https://huggingface.co/Nateemerson/DeepSeek-V4.1-Flash-1017-EXL3-3bpw-2x-RTX-PRO-6000

Nate’s in SF @NateemersonOriginally posted 2026.10.18
Open original on X
x.com/nateemerson/status/2111693164704944222Archived 2026.10.18 16:24
Open original on X