The base model bridges the gap between Google's open-weights Gemma 4 architecture and Claude 4.6's profound reasoning capabilities, fine-tuned using high-effort reasoning datasets.

Stop Sequences

When running this model, ensure you configure the following stop sequences in your inference engine to prevent endless generation:

<end_of_turn>
<eos>

Provided Files

This repository includes the following quantizations to suit various memory and compute requirements, along with the required multimodal projector files:

Model Files:

gemma-4-31B-it-Claude-Opus-Distill.Q3_K_M.gguf
gemma-4-31B-it-Claude-Opus-Distill.Q4_K_M.gguf
gemma-4-31B-it-Claude-Opus-Distill.Q5_K_M.gguf
gemma-4-31B-it-Claude-Opus-Distill.Q6_K.gguf
gemma-4-31B-it-Claude-Opus-Distill.Q8_0.gguf

Projector Files:

mmproj-BF16.gguf
mmproj-F16.gguf
mmproj-F32.gguf

Core Capabilities

Optimized through high-effort reasoning distillation datasets from Claude Opus interactions, this model excels in:

Coding: Advanced code generation, debugging, and software architecture planning.
Science: Deep scientific reasoning, hypothesis evaluation, and analytical problem-solving.
Deep Research: Navigating complex, multi-step research queries and synthesizing vast amounts of information.
General Purpose: Highly capable instruction-following for everyday tasks requiring high logical coherence.

Acknowledgements & Credits

TeichAI: For fine-tuning and developing the original model architecture and datasets.
Google: For providing the foundational Gemma 4 - 31B open weights model.
Unsloth: For the memory and compute-optimized fine-tuning framework used to train the base model.
Crownelius: For contributing structured Opus reasoning datasets.

Original Model: TeichAI/gemma-4-31B-it-Claude-Opus-Distill

Abiray/gemma-4-31B-it-Claude-Opus-Distill-GGUF

Gemma 4 - 31B x Claude Opus 4.6 (GGUF)

Stop Sequences

Provided Files

Core Capabilities

Acknowledgements & Credits

Abiray/
gemma-4-31B-it-Claude-Opus-Distill-GGUF