Downloads · 30 days
13
14% of all-time downloads
MiniCPM-o-4_5-gguf
orca-zhang/MiniCPM-o-4_5-gguf
MiniCPM-o-4_5-gguf is a machine learning model from orca-zhang. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as apache-2.0.
This repository contains a Q6K mixed-precision vision projector for the official MiniCPM-o 4.5 GGUF model.
All-time downloads
90
Public
Repo size
732 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.gguf732 MB · 100%
At a glance
- License
- apache-2.0
- Access
- Public
- Created
- Aug 26, 2026
- Updated
- Aug 26, 2026
- SHA
- 07775f34
Base models
Model card
From the Hugging Face model README
MiniCPM-o 4.5 Q6 Vision Projector
This repository contains a Q6_K mixed-precision vision projector for the official MiniCPM-o 4.5 GGUF model.
Artifact
- File:
vision/MiniCPM-o-4_5-vision-Q6_K.gguf - Size:
731856256bytes - SHA256:
4e30763cb9d23f86c4ae07034f72b124c4855c7f4477b61fcf5dac66b76930a1 - Source projector SHA256:
1453678cc4e4fe18de241952962e234f265cb8dda780773526103ab8ba82f421 - Quantizer: llama.cpp build
10615, revisionf280b2698 - Quantization:
Q6_K; tensors unsupported by K-quant retain the quantizer's safe F16/Q8 fallback types. Effective size is about 11.11 bits per weight.
RTX 3050 Validation
Validated with MiniCPM-o 4.5 Q4_K_M, CUDA llama-server build 10615, one slot, Q8 KV cache, 1536-token context, Flash Attention, and 768x768 input.
- 37/37 language-model layers offloaded to CUDA
- 48/48 difficult-image runs completed
- 0 runtime errors, parse errors, truncations, or OOM events
- Mean 5.07 seconds/image; P95 5.64 seconds/image
- Mean decode 29.36 tokens/second
- GPU memory 5505 MiB on a 6 GiB RTX 3050
The original MiniCPM license and model terms continue to apply.
- License
- apache-2.0
- Created
- Aug 26, 2026
- Updated
- Aug 26, 2026