G4-MeroMero-v2-31B-Q6_K_L-GGUF

Llamacpp imatrix Quantization of G4-MeroMero-v2-31B by <a href="https://huggingface.co/zerofata">zerofata</a>

Using <a href="https://github.com/ggml-org/llama.cpp/">llama.cpp</a> release <a href="https://github.com/ggml-org/llama.cpp/releases/tag/b10516">b10516</a> for quantization.

Original model: https://huggingface.co/zerofata/G4-MeroMero-v2-31B

Quant details

• This is a 7.05bpw quant that follows <a href="https://huggingface.co/bartowski">bartowski</a>'s Gemma 4 31b Q6_K_L quantization scheme. Among other things, it uses Q8_0 for embed and output weights.
• The following <a href="https://huggingface.co/zerofata/G4-MeroMero-v2-31B-GGUF/blob/main/G4-MeroMero-v2-31B-imatrix.dat">imatrix</a> was used to create this quant, provided by the model's creator, <a href="https://huggingface.co/zerofata">zerofata</a>.
• mmproj and mtp meant for default Gemma 4 31b instruct work properly with this model and were also uploaded to this repository. They were made by <a href="https://huggingface.co/bartowski">bartowski</a> and taken from his <a href="https://huggingface.co/bartowski/google_gemma-4-31B-it-GGUF">gemma-4-31B-it-GGUF</a> repository.

Visit the original model page https://huggingface.co/zerofata/G4-MeroMero-v2-31B for more information, including recommended samplers and premade text completion presets for SillyTavern.