g023/NeuronBlade-Qwen3.5-4B

Model Card for Model ID

So this model is made by using (https://github.com/g023/neuronblade) to alter a Qwen3.5 4B model to improve effectiveness. It is a WIP so feel free to test with me. I will be updating as I go. MODEL CAN HAVE UNPREDICTABLE OUTPUTS

llama-cli -m NeuronBlade-Qwen3.5-4B-Q4_K_M.gguf -n 8192 --temp 1.0 --top-p 0.9 --repeat-last-n 16384 --mirostat 2 --mirostat-lr 0.2 --mirostat-ent 3 --presence-penalty 0.3 --frequency-penalty 0.5 --repeat-penalty 0.4