Qwen3.5-2B-fullRC GGUF

GGUF conversion of RockMan256/Qwen3.5-2B-fullRC.

Files

| File | Description | Size |

|------|-------------|------|

| qwen3.5-2B-fullRC-Q4_K_M.gguf | LLM weights in Q4_K_M | 1.24 GB |

| mmproj-model-f16.gguf | Vision projector (mmproj) in F16 | 668 MB |

Usage with llama.cpp

# Build llama.cpp (if not already built)
cd llama.cpp && make -j

# Run with multimodal (image + text) inference
./llama-mtmd-cli \
  -m qwen3.5-2B-fullRC-Q4_K_M.gguf \
  --mmproj mmproj-model-f16.gguf \
  --chat-template chatml

# Text-only inference (no vision)
echo "What is the capital of France?" | ./llama-mtmd-cli \
  -m qwen3.5-2B-fullRC-Q4_K_M.gguf \
  -p - --chat-template chatml

Architecture

• Text: Qwen3.5 (hybrid linear + full attention, 24 layers, 2B params)
• Vision: ViT encoder (24 layers, 1024 hidden)
• Vocab: 248,320 tokens
• Context: 262,144 tokens

License

Apache 2.0 (same as base model)