Hob-forge/Kolibri-1-GGUF community quant repo

Community GGUF of Aleph-Alpha/Kolibri-1, not an Aleph Alpha release. Read 2026-10-17. Architecture tag kolibri1, 78B. Published file sizes: Q2_K 28.6 GB, Q3_K_M 37.5 GB, Q4_K_M 47.5 GB, Q5_K_M 55.6 GB, Q6_K 64.2 GB, Q8_0 83.1 GB. The README reports llama.cpp measurements, including Q4_K_M with cpu-moe on an RTX 5070 (12 GB) keeping routed experts in system RAM and using 6.4 GB VRAM at 262,144 context, plus a long-context recall test. It explicitly says quality versus the original FP8 model was not benchmarked, so the file sizes are a runnability source and not a quality source. Requires a llama.cpp build that knows the kolibri1 arch.