tianrui6641/gemma-4-31b-gguf-mxfp4

Gemma 4 31B GGUF MXFP4

This repo contains a GGUF export of google/gemma-4-31B quantized to dense MXFP4, plus the matching mmproj GGUF for image input support.

Files

• gemma-4-31B.MXFP4.gguf: main text model quantized to dense MXFP4.
• mmproj-gemma-4-31B.f16.gguf: multimodal projector required for image input.

LM Studio note

The final GGUF now embeds a Gemma 4-compatible chat template that forces thinking on, even in runtimes that pass enable_thinking=false or do not expose a separate reasoning toggle.

This repo also ships tokenizer_config.json and chat_template.jinja sidecars with the Gemma 4 response_schema for the <|channel>thought\n...<channel|> reasoning block, so frontends that look beyond GGUF metadata can both elicit and parse reasoning more reliably.

This is the base google/gemma-4-31B checkpoint, not google/gemma-4-31B-it. The reasoning-aware, tool-capable template is embedded so the runtime keeps both tool formatting and thinking support, but instruction-following and tool-calling quality may differ from the -it model.