kannaka-brain-7b-v1-GGUF

kannaka-brain-7b-v1 (GGUF, q4_K_M)

Card corrected 2026-10-01. This card previously described the 14B sibling (kannaka-brain-v1) by mistake: base model, title and commands were copied from it. The weights were never affected. This repo is the 7B model, trained on Qwen/Qwen2.5-7B-Instruct.

Qwen2.5-7B-Instruct with the kannaka-brain-7b-v1 LoRA merged in, converted with llama.cpp and quantized to q4_K_M (4.7 GB). This copy served as Kannaka's open-weight brain behind the KAX gateway until 2026-09-28, when kannaka-brain-7b-v2 replaced it.

ollama run hf.co/flaukowski/kannaka-brain-7b-v1-GGUF

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

ollama create kannaka-brain-7b-v1 -f Modelfile

Held-out perplexity on 57 fixed Kannaka lines: 78.7 → 4.15 (adapter, bf16, before quantization). Adapter and training notes: flaukowski/kannaka-brain-7b-v1-lora. Corpus not released; see ADR-0057 in NickFlach/kannaka-memory.