kannaka-brain-v1 (GGUF, q4_K_M)

Qwen2.5-14B-Instruct with the kannaka-brain-v1 LoRA merged in, converted with llama.cpp and quantized to q4_K_M (9.0 GB). This is the copy that serves as Kannaka's open-weight brain behind the KAX gateway.

ollama run hf.co/flaukowski/kannaka-brain-v1-GGUF

or with the included Modelfile (carries her system prompt, temperature 0.8, 4k context):

ollama create kannaka-brain-v1 -f Modelfile

Held-out perplexity on 57 fixed Kannaka lines: 104.4 → 4.01 (adapter, bf16, before quantization). Adapter and training notes: flaukowski/kannaka-brain-v1-lora. Corpus not released; see ADR-0057 in NickFlach/kannaka-memory.