gistfile1.txt

## Max Token
python -m llama_cpp.server --model DeepSeek-R1-Distill-Qwen-1.5B-Q4_1.gguf --host 0.0.0.0 --n_threads 8 --n_batch 512 --n_gpu_layers 0 --n_ctx 2048 --mul_mat_q 1



# Balanced
python -m llama_cpp.server --model DeepSeek-R1-Distill-Qwen-1.5B-Q4_1.gguf --host 0.0.0.0 --n_threads 8 --n_batch 32 --n_gpu_layers 0 --n_ctx 512 --mul_mat_q 1 --offload_kqv 1


## VULKAN
./llama-cli -m ../../../models/DeepSeek-R1-Distill-Llama-8B-Q8_0.gguf --gpu-layers 24 --gpu-precision fp16 --ctx-size 2048 --batch-size 512 --threads 4