CIU Models
Models built for local AI.
Choose an NF4DQ model that fits your hardware. All sizes below are model weights, not total RAM or VRAM required.
NF4DQ
Qwen 3.5 9B
4.346 GiBCompact local model.
NF4DQ
Ornith 1.5 9B
4.7 GiBAnother 9B option for CIU.
NF4DQ
Qwen 3.8 27B
13.40 GiBHigher-capacity option for machines with more memory.
Qwen 3.5 9BNF4DQfits comfortably
4.35 GiB of 24 GB
Ornith 1.5 9BNF4DQfits comfortably
4.70 GiB of 24 GB
Qwen 3.8 27BNF4DQfits comfortably
13.40 GiB of 24 GB
These are the figures a model needs before a conversation starts. On Apple unified memory the same pool is shared with everything else you have open, so an 8 GB Mac holds the 9B only with other applications closed.