News for hackers. Stuff that matters.
Re:Kimi K3.1 Pro weights out by user_7832 on Wednesday November 04, 2026 @01:09PM (#50369320)
Fwiw, the Qwen 3.8 27B at 4 bit quant is apparently quite good and is probably the best recent model. It runs (iirc) on 18-22ish GB of memory.
Re:Kimi K3.1 Pro weights out by robertkarl on Wednesday November 04, 2026 @01:30PM (#50369620)
quantization has come a really long way. Q3 is fine, especially GSQ-RCO. You can run 27B well on 16GB of VRAM.
Re:Kimi K3.1 Pro weights out by zozbot234 on Wednesday November 04, 2026 @02:10PM (#50370143)
The point of quantizing larger models is to beat other models running on the same limited-VRAM platform, i.e. 16GB VRAM in your example. You're doing "well" compared to other models meant for running on 16GB VRAM; you're not doing "well" at all compared to those running higher precisions for the same params up to at least Q8. (16 bit and higher is where the gain from further precision gets real iffy.)
Re:Kimi K3.1 Pro weights out by adornKey on Wednesday November 04, 2026 @02:59PM (#50370711)
With e.g. Strata you can run 125B Q4 very well on 16 GB of VRAM. 27B already has vintage vibes... OH, the good old 27B days...
Re:Kimi K3.1 Pro weights out by ricardobeat on Wednesday November 04, 2026 @04:44PM (#50371981)
I read somewhere that Qwen 4 is right around the corner. Praise be!