Mattimax/DAC6.5-GGUF

This repository contains the official GGUF quantized variants of DAC6.5, the flagship multimodal model by M.INC. built on our custom neural architecture series.

These quantized builds are optimized for lightweight, low-latency, and memory-efficient local deployment using GGUF-compatible inference runtimes (such as llama.cpp and Ollama).

---

Base Model & Architecture Overview

DAC6.5 unifies vision perception and language understanding in a compact, single-pipeline system:

• Base Model Repository: Mattimax/DAC6.5
• Language Backbone: LFM2.5-230M
• Vision Encoder: SigLIP2
• Projector Layer: Proprietary cross-modal projection layer trained by M.INC.

---

Provided Quantization Formats

This repository provides two high-efficiency GGUF quantization formats:

• dac6.5-q4_k_m.gguf: 4-bit K-quantized weights balancing fast execution, minimal memory usage, and high retention of multimodal performance. Ideal for edge devices, single-board computers, and low-VRAM environments.
• dac6.5-q8_0.gguf: 8-bit quantized weights offering maximum precision, near-lossless accuracy relative to the FP16 base model, and robust stability for critical evaluation workloads.

---

☕ Support my research

---

About M.INC.

M.INC. focuses on research and development of custom neural architectures, efficient language models, and accessible multimodal systems. DAC6.5 represents the first milestone in our custom architecture series, establishing a new baseline for compact, locally deployable artificial intelligence.