Downloads · 30 days
140
9% of all-time downloads
dylanjkl/JKL-Luau-Gemma-4-31B-it-Claude-Opus-Distill-GGUF
JKL-Luau-Gemma-4-31B-it-Claude-Opus-Distill-GGUF is a text generation model from dylanjkl. Use it when you need the model to write or continue text. The card lists the license as apache-2.0.
This repository contains the GGUF quantized formats of JKL-Luau-Gemma-4-31B-it-Claude-Opus-Distill-GGUF.
Downloads · 30 days
140
9% of all-time downloads
All-time downloads
1.6K
Public
Repo size
51.3 GB
Likes
2
Public
Click a slice to open those files.
.gguf51.3 GB · 100%
From the Hugging Face model README
This repository contains the GGUF quantized formats of JKL-Luau-Gemma-4-31B-it-Claude-Opus-Distill-GGUF.
This model is a highly specialized fine-tune of Gemma-4-31B-it focused extensively on the Luau programming language (native to Roblox). It was developed using Quantization-Aware Distillation (QAD) methodologies and distilled using high-quality coding trajectories generated from Claude Opus, providing state-of-the-art zero-shot capabilities for Luau game development, bug resolution, and architectural design.
Task library semantics, and Modern UI (e.g., GuiObject layout hierarchies).llama.cpp, text-generation-webui, and Ollama.JKL-Luau-Gemma-4-31B-it-Claude-Opus-Distill.Q8.gguf (8-bit quantization, great quality and great performance).JKL-Luau-Gemma-4-31B-it-Claude-Opus-Distill.Q4_K_M.gguf (4-bit quantization suitable for machines with limited VRAM).llama.cppYou can run the model directly inside a local environment using llama.cpp:
# Example using the Q8_0 quant
./main -m JKL-Luau-Gemma-4-31B-it-Claude-Opus-Distill.Q8_0.gguf \
--color \
-c 32768 \
--temp 0.6 \
-p "<bos><start_of_turn>user\nWrite a robust Luau character sprinting script with Server/Client validation.<end_of_turn>\n<start_of_turn>model\n"
Note: Ensure your context window (-c) is set appropriately for your available VRAM, as this model supports Gemma 4's extended context length.
loadstring without appropriate flags, or specific os functions).Fine-tuned on the NVIDIA RTX Pro 6000 Blackwell workstation using the Unsloth library.
1x RTX Pro 6000 Blackwell 96GB Workstation Edition over the course of 20 hours with a high LoRA rank of 32 for enhanced logical throughput.This model inherits the license of the base model [Google Gemma-4] and follows responsible AI distillation guidelines from Anthropic.