WebBrain Compass VLD v6 (2B) GGUF

GGUF quantizations of webbrain-one/webbrain-compass-vld-v6-2b for local and in-browser execution with llama.cpp and WebGPU.

Included Files

• webbrain-compass-vld-v6-2b-Q4_K_M.gguf (1.1 GB): Default compact quantization for browser runtime.
• webbrain-compass-vld-v6-2b-Q8_0.gguf (1.8 GB): High precision quantization.
• webbrain-compass-vld-v6-2b-BF16.gguf (3.3 GB): Unquantized BF16 weights.
• mmproj-webbrain-compass-vld-v6-2b-F16.gguf (362 MB): Vision multimodal projector weights.

Usage with llama-server

llama-server \
  --model webbrain-compass-vld-v6-2b-Q4_K_M.gguf \
  --mmproj mmproj-webbrain-compass-vld-v6-2b-F16.gguf \
  --alias webbrain-compass-vld-v6-2b \
  --ctx-size 16384 \
  --jinja \
  --reasoning off \
  --port 8080