Hugging Peers

MiniCPM-V-4.7-35B-A3B-APEX-I-NanoPlus-GGUF

☆ Save
by IsValorum image-text-to-text gguf apache-2.0
2.1K downloads 7 likes on Hugging Face updated

Model card

Quick Navigation Index

Support on Ko-fi

Independently computed using dedicated cloud compute clusters, funded 100% out-of-pocket.

If you consider these releases to be of high quality and they help you in your day-to-day workflow, consider supporting on Ko-fi. Every contribution directly sustains future cloud compute hours and helps support my university studies.

MiniCPM-V 4.7 35B-A3B APEX-I NanoPlus V2.1 GGUF

Official APEX-I NanoPlus V2.1 canonical release for OpenBMB MiniCPM-V 4.7 35B-A3B, engineered for ultra-compact on-device memory footprints while preserving recurrent DeltaNet dynamics and visual fidelity.

Key Architectural Features:

  • Canonical 733-Tensor Allocation:
    • output.weight in Q6_K and token_embd.weight in Q4_K.
    • MoE Expert Routers (ffn_gate_inp) in F32.
    • Shared foundation experts (shexp) in Q4_K.
    • Recurrent SSM continuous states in F32 (ssm_a, ssm_conv1d, ssm_dt, ssm_norm).
    • Border down projections in Q3_K and border gate/up in IQ3_XXS.
    • Core down projections (layers 2-37) in IQ3_XXS, gate in IQ2_S, protected up (layers 2-15) in IQ2_S, compact up (layers 16-37) in IQ2_XXS.
  • Multimodal Projector:
    • Includes companion mmproj-Q8_0.gguf for on-device visual inference.

CLI Usage Example:

llama-minicpmv-cli \
  -m MiniCPM-V-4.7-35B-A3B.APEX-I-NanoPlus-V2.1.gguf \
  --mmproj mmproj-Q8_0.gguf \
  --image test.jpg \
  -p "Describe what you see in this image in detail." \
  -c 4096 --temp 0.7