models:
  gemma-4-12b:
    gpu_url: http://192.168.68.110:8080/v1
    sidecar_url: http://192.168.68.110:8090
    gpu_host: 192.168.68.110
    label: Gemma-4 12B (RTX 5070)
    max_concurrent: 2
    context: 262144
    tiers: [starter, professional, enterprise]
    capabilities: [completion, multimodal]
    model_path: /home/llmuser/models/gemma-4/gemma-4-12b-it-Q4_K_M.gguf
    args: --mmproj /home/llmuser/models/gemma-4/mmproj-F16.gguf --spec-draft-model /home/llmuser/models/gemma-4/gemma-4-12b-it-Q8_0-MTP.gguf --spec-type draft-mtp --spec-draft-n-max 4 --batch-size 2048 --ubatch-size 4096 --n-gpu-layers 99 --flash-attn 1 --image-min-tokens 2048 --image-max-tokens 16384

  qwen3.6-27B-code:
    gpu_url: http://192.168.68.8:8080/v1
    sidecar_url: http://192.168.68.8:8090
    gpu_host: 192.168.68.8
    label: Qwen3.6 27B Code (RTX 3090)
    max_concurrent: 2
    context: 262144
    tiers: [professional, enterprise]
    capabilities: [completion]
    model_path: /root/models/Qwen3.6-27B-NEO-CODE-2T-OT-IQ4_NL.gguf
    args: -ctk turbo4 -ctv turbo4 --flash-attn on --reasoning off --spec-type draft-mtp --spec-draft-n-max 2 -t 8

  ornith-1.0-35b:
    gpu_url: http://192.168.68.15:8080/v1
    sidecar_url: http://192.168.68.15:8090
    gpu_host: 192.168.68.15
    label: Ornith-1.0 35B (Strix Halo)
    max_concurrent: 1
    context: 4096
    tiers: [professional, enterprise]
    capabilities: [completion]
    model_path: /data/llamaccp-models/Qwen3.6-35B-A3B-Opus-IQ4_NL.gguf
    args: -c 262144 -ngl 99 --flash-attn on

hosts:
  gpu-light:
    address: 192.168.68.110
    gpu_name: NVIDIA GeForce RTX 5070
    vram_gb: 12
    current_model: gemma-4-12b

  gpu-dense:
    address: 192.168.68.8
    gpu_name: NVIDIA GeForce RTX 3090
    vram_gb: 24
    current_model: qwen3.6-27B-code

  gpu-moe:
    address: 192.168.68.15
    gpu_name: AMD Strix Halo (iGPU)
    vram_gb: 64
    current_model: ornith-1.0-35b