spiritfather/Melinoe-Qwen3-8-27B-VL-i1-GGUF

About

weighted/imatrix quants of https://huggingface.co/bgg1996/Melinoe-Qwen3-8-27B-VL

Benchmarked on CaliperBench — a creative-writing benchmark scoring prose craft, roleplay and willingness rather than general intelligence. See this model's scores: caliperbench.com/m/melinoe-qwen3.8-27b-vl.

Static (non-imatrix) quants of this model are at spiritfather/Melinoe-Qwen3-8-27B-VL-GGUF.

imatrix (importance matrix) computed over bartowski calibration_datav3. Weighted quants spend precision where it matters most — the low bit-rates (IQ1–IQ4) are meaningfully better than same-size static quants; at Q5/Q6 the difference is negligible.

Vision: this is a VL model. llama.cpp keeps the vision tower OUT of the main GGUF, so multimodal use needs mmproj-Melinoe-Qwen3-8-27B-VL-BF16.gguf alongside your chosen quant (llama-server -m <quant>.gguf --mmproj mmproj-Melinoe-Qwen3-8-27B-VL-BF16.gguf). Text-only use needs just the quant.

Usage

If you are unsure how to use GGUF files, refer to one of TheBloke's READMEs for more details, including on how to concatenate multi-part files.

Provided Quants

(sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)