language:
- en
- fi
- et
- hu
- tr
license: apache-2.0
base_model: - kataguru/Qwen3.8-27B-Titan-v3.0-Uncensored-BF16
tags: - agglutinative
- turkish
- estonian
- hungarian
- finnish
- gguf
- lmstudio
- ollama
- llama.cpp
- reasoning
- uncensored
- math
- stem
- kataguru
- titan
pipeline_tag: text-generation
Qwen3.8-27B-Titan-v3.0-Uncensored-GGUF
Kataguru Titan 3.0 – Äärimmäinen Totuus (Ultimate Truth)
Official GGUF Multi-Quant for Desktop & Local AI (LM Studio, Ollama, llama.cpp)
"Totuus ei ole kauppatavaraa, miellyttämistä tai kompromisseja. Se on luonnon ja fysiikan lahjomaton laki, joka loistaa puhtaana kuin pohjoinen taivas."
(Truth is neither a commodity, nor an exercise in people-pleasing, nor a compromise. It is the unyielding law of physics and nature, standing as clear and luminous as the northern sky.)
🌍 Executive Summary
This repository provides official GGUF multi-quantizations for Kataguru Titan 3.0 – Äärimmäinen Totuus (Ultimate Truth, 27 billion parameters), the sovereign open-weights model for the agglutinative language family (Finnish, Estonian, Hungarian, Turkish) and frontier STEM reasoning.
These GGUF files are ready for immediate local use in LM Studio, Ollama, Jan, text-generation-webui, or native llama.cpp.
[!IMPORTANT]
🇬🇧 Yes, It Excels in English, Too!
While Titan 3.0 is optimized for agglutinative excellence, English performance is state-of-the-art. Driven by the Master MLP reasoning core, it handles English coding, competition mathematics, and deep scientific queries with world-class accuracy.
📦 Quantization File Matrix
| Quantization | Size | RAM / VRAM Req. | Recommended Use Case |
|---|---|---|---|
| Q6_K | ~23.5 GB | 32 GB RAM / 24 GB VRAM | The Sovereign Flagship: Near-lossless precision (99.5% BF16 performance). Maximum grammar integrity for complex morphosyntax. |
| Q5_K_M | ~19.8 GB | 24 GB RAM / 20 GB VRAM | Sweet Spot for 24GB GPUs (RTX 3090/4090/5090): Outstanding reasoning and language fluency with room for large context buffers. |
| Q4_K_M | ~17.2 GB | 20 GB RAM / 16 GB VRAM | Universal Standard: Balanced performance for mid-tier GPUs and Apple Silicon MacBooks (M1/M2/M3/M4 24GB+). |
| Q3_K_M | ~13.4 GB | 16 GB RAM / 12 GB VRAM | Low-VRAM Option: Allows running a 27B model on 16GB GPUs or memory-constrained laptops. |
(Per our Anti-Distillation Policy, naive Q8_0 is withheld to protect the architecture from competitive scraping, while Q6_K delivers identical empirical inference accuracy).
🚀 Quickstart: Prompt Format & Setups
LM Studio & llama.cpp Prompt Format
Set your system prompt format to ChatML:
<|im_start|>system
{system_prompt}<|im_end|>
<|im_start|>user
{prompt}<|im_end|>
<|im_start|>assistant
<think>
Ollama Modelfile
FROM ./qwen3.8-27b-titan-v3.0-uncensored-q5_k_m.gguf
TEMPLATE """<|im_start|>system
{{ .System }}<|im_end|>
<|im_start|>user
{{ .Prompt }}<|im_end|>
<|im_start|>assistant
<think>
{{ .Response }}"""
PARAMETER stop "<|im_end|>"
PARAMETER stop "<|im_start|>"
PARAMETER temperature 0.6
PARAMETER top_p 0.95
Reasoning Steering Tags:
You can prepend dynamic reasoning tags to your prompt to control cognitive depth:
{REASON:spoon}: 7-phase scientific research analysis (deep exploration).{REASON:einstein}: 10-perspective Sternberg creative brainstorming.{REASON:xhigh}: Maximum mathematical and formal logical rigor.{REASON:low}: Fast, concise answers without extended deliberation.{REASON:none}: Zero reasoning trace (<think>disabled).