CoolFace
Modelpublic

Abiray/gemma-4-31B-Claude-4.6-Opus-Reasoning-Distilled-GGUF

sourceHugging Faceupdated 6mo agoView on Hugging Face
11likes815downloads
Model Card

gemma-4-31B-Claude-4.6-Opus-Reasoning-Distilled-GGUF

This repository contains GGUF format model files for EganAI/gemma-4-31B-Claude-4.6-Opus-Reasoning-Distilled.

Model Details

  • —Base Architecture: Gemma 4 (31B parameters)
  • —Training Focus: Full parameter SFT on 12,680 Claude Opus 4.6 reasoning traces.

Available Quantizations

FileSize
gemma-4-31B-Claude-4.6-Opus-Reasoning-Distilled-Q3_K_M.gguf15.3 GB
gemma-4-31B-Claude-4.6-Opus-Reasoning-Distilled-Q4_K_M.gguf18.7 GB
gemma-4-31B-Claude-4.6-Opus-Reasoning-Distilled-Q5_K_M.gguf21.8 GB
gemma-4-31B-Claude-4.6-Opus-Reasoning-Distilled-Q6_K.gguf25.2 GB
gemma-4-31B-Claude-4.6-Opus-Reasoning-Distilled-Q8_0.gguf32.6 GB

Recommendation: Q4_K_M provides the optimal balance between inference speed, memory consumption, and preserving the model's reasoning accuracy.

Stop Sequence

To ensure generation stops cleanly, configure your inference engine or UI to use the following stop sequence (native to the Gemma 4 template):

  • —<end_of_turn>

Usage Instructions

Using llama.cpp CLI

bash
./llama-cli -m gemma-4-31B-Claude-4.6-Opus-Reasoning-Distilled-Q4_K_M.gguf -p "Prove that the square root of 2 is irrational." -n 1024