Inkling-2-Mini in 5.4 GB Memory
129 comments
- Points424
- Comments107
129 comments
Story details from Hacker News ↗ · linked article on github.com.
Inkling-2-Mini in 5.4 GB Memory was first seen on Hacker News · Front page on Nov 21, 2026. Since then it has appeared in 93 checks on 2 different days. Its best position was #1 of 30, reached on Nov 21, 18:59 UTC. It is #18 right now, slipping from #14 to #18 over the last 6 checks.
Position on Hacker News · Front page at every check (lower is better).
Best rank per day (UTC).
points as reported by Hacker News.
A few top-level comments from Hacker News, shortened. Read the full thread on Hacker News.
If you're new to quantized models: Don't get too excited about claims of fitting large models into tiny RAM footprints from random GitHub repos. There is always a tradeoff, and it's too easy to benchmaxx KL-divergence numbers and even single benchmark scores. Just because you can get something to run on a device…
Our future hinges on the capability gap between frontier lab models and open source local models
I had a rough time with llama.cpp so fixed the bugs in it for performance some notes you need a minimum of 12gb of vram for credible performance with decent context length above 100k... The best I could do with established techniques and staying at nvfp4 levels of quant, is around 40-45 toks/s on 5070 Ti with 800…
Archived daily, weekly and monthly lists that include this entry.
Yes. As of Nov 22, 17:29 UTC it is #18, slipping from #14 to #18 over the last 6 checks.
#1 on Hacker News · Front page, on Nov 21, 18:59 UTC. Trendoday checks this list every 15 minutes.
It was first seen on Nov 21, 2026 and has been on the list on 2 days (93 checks) since.
424 points and 107 comments; submitted by k__ on Nov 21, 2026.