To put this on perspective, consider that with 128GB of unified RAM it can run even fast: https://x.com/antirez/status/2119392658661749175
Like, coding-agent acceptably fast. See later in the reply for the prefill at ~300 t/s but can go even faster for larger prefills.
probably; 32-64 is the range where local models get extraordinarily useful, and in my experience any sub-4 quants get ridiculously incoherent, which is why those models tend to try to 'wow' people with tok/s rather than benchmarks.
This is quite an obnoxious Readme.md with AI slop. It wraps around llama.cpp but without seeing what platform assumptions it makes its unclear for other readers whether this is worth testing.
EDIT: I think it's 2 bit? If so, would be curious to see the quality of the output. Not great, I think.
haven't tried it yet though.