mradermacher

/

GLM-4.6-REAP-268B-A32B-Derestricted-i1-GGUF

like 5

Getting ~15 tps output on my M2 Ultra 192GB. This model is surprisingly stable and intelligent for having been partially lobotomized! Multi-turn tool-calling works in Kilo Code and LM Studio without a hitch. Even 65K long-context is holding very well (going to try for double and maybe even quadruple context later on).

Hope you're able to reap a GLM-4.7 at some point!