Modelpublic

liodon-ai/MaralGPT-Mythos-9B-2606-GGUF-imatrix-GGUF

sourceHugging Faceotherupdated 3mo agoView on Hugging Face

0likes199downloads

Model Card

MaralGPT-Mythos-9B-2606-GGUF — iMatrix GGUF

GGUF quantizations of MaralGPT/MaralGPT-Mythos-9B-2606-GGUF, published by Liodon AI.

Quick Start

llama.cpp

bash

llama-cli -hf liodon-ai/MaralGPT-Mythos-9B-2606-GGUF-imatrix-GGUF:Q4_K_M

Ollama

bash

ollama run hf.co/liodon-ai/MaralGPT-Mythos-9B-2606-GGUF-imatrix-GGUF:Q4_K_M

LM Studio / Jan — search liodon-ai/MaralGPT-Mythos-9B-2606-GGUF-imatrix-GGUF and pick your quant.

Quants

Quant	Size	VRAM est.	Notes
`IQ2_M`	3.61 GB	~4 GB	2-bit, iMatrix — smallest usable
`IQ3_M`	4.42 GB	~5 GB	3-bit, iMatrix — great quality/size tradeoff
`IQ4_XS`	5.20 GB	~6 GB	4-bit extra-small, iMatrix
`Q4_K_M`	5.63 GB	~6 GB	4-bit, iMatrix-calibrated (recommended)
`Q5_K_M`	6.47 GB	~7 GB	5-bit, iMatrix-calibrated
`Q6_K`	7.36 GB	~8 GB	6-bit, iMatrix-calibrated, near-lossless
`Q8_0`	9.53 GB	~11 GB	8-bit, essentially lossless

What is iMatrix?

Standard quantization treats all weights equally. iMatrix runs 128 calibration chunks through the full-precision model to find which weights matter most, then allocates more precision where it counts. At Q2/Q3/Q4 this means noticeably better coherence and instruction-following — same file size, better output.

Calibration: 2M tokens of WikiText-103.

Also see plain (non-iMatrix) quants: liodon-ai/MaralGPT-Mythos-9B-2606-GGUF-GGUF

Source

—Model: MaralGPT/MaralGPT-Mythos-9B-2606-GGUF
—License: other

Quantized by [Liodon AI](https://huggingface.co/liodon-ai)