Ornstein-27B-GGUF
GGUF quantizations of DJLougen/Ornstein-27B — a reasoning-focused fine-tune of Qwen 3.5 27B trained on 1,229 high-quality reasoning traces curated through a custom Drift Diffusion Modeling (DDM) pipeline.
Support This Work
I'm a PhD student in visual neuroscience at the University of Toronto who also happens to spend way too much time fine-tuning, merging, and quantizing open-weight models on rented H100s and a local DGX Spark. All training compute is self-funded — balancing GPU costs against a student budget. If my uploads have been useful to you, consider buying a PhD student a coffee. It goes a long way toward keeping these experiments running.
What Makes Ornstein Different
Unlike typical reasoning fine-tunes that use large volumes of synthetic data, Ornstein implements quality-over-quantity:
- Detects degenerate reasoning: Identifies "fake" reasoning that mimics thought without substance (hedging, restating, circling)
- Premium vs. Degenerate split: 799 premium traces + 430 selected degenerate traces = 1,229 total
- DDM AUC of 0.9705 separating premium from degenerate reasoning with 99.49% sensitivity
The model uses <think>...</think> blocks for extended multi-phase reasoning with self-correction and verification before providing final answers.
Available Quantizations
| Quantization | Size | Use Case |
|---|---|---|
| F16 | 53.8 GB | Full precision, no quality loss |
| Q8_0 | 28.6 GB | Near-lossless, good for high-end consumer GPUs |
| Q6_K | 22.1 GB | High quality |
| Q5_K_M | 19.2 GB | Good balance |
| Q5_K_S | 18.7 GB | Lighter variant |
| Q4_K_M | 16.5 GB | Recommended — strong quality/size tradeoff |
| IQ4_XS | 11.6 GB | Efficient 4-bit |
| Q3_K_L | 14.3 GB | Lighter 3-bit |
| Q3_K_M | 13.3 GB | Mid 3-bit |
| Q3_K_S | 12.1 GB | Light 3-bit |
| Q2_K | 10.7 GB | Minimal footprint |
Quick Start
llama.cpp
# Download a quantization (example: Q4_K_M)
huggingface-cli download DJLougen/Ornstein-27B-GGUF ornstein-27b-q4_k_m.gguf --local-dir .
# Run with llama.cpp
./llama-cli -m ornstein-27b-q4_k_m.gguf \
-p "You are a helpful reasoning assistant." \
--temp 0.6 -n 8192
Ollama
# Create a Modelfile
cat <<EOF > Modelfile
FROM ./ornstein-27b-q4_k_m.gguf
PARAMETER temperature 0.6
PARAMETER num_predict 8192
SYSTEM "You are a helpful reasoning assistant."
EOF
ollama create ornstein -f Modelfile
ollama run ornstein
LM Studio
- Download the desired quantization from the Files tab
- Load it in LM Studio
- Set context length to 8192 for full reasoning depth
Recommended Settings
| Parameter | Suggested Value |
|---|---|
| Temperature | 0.6 |
| Top-P | 0.95 |
| Max Tokens | 8192 |
| Repeat Penalty | 1.1 |
Training Details
| Parameter | Value |
|---|---|
| Base Model | unsloth/Qwen3.5-27B |
| Parameters | 27B |
| Method | LoRA (rank 32, alpha 32) |
| Dropout | 0.05 |
| Epochs | 1 |
| Learning Rate | 1e-4 (cosine schedule, 10% warmup) |
| Max Sequence Length | 8192 |
| Micro Batch Size | 1 |
| Gradient Accumulation | 4 steps |
| Weight Decay | 0.01 |
| LoRA Targets | q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj |
| Framework | Unsloth |
Data Quality Metrics
| Metric | Value |
|---|---|
| Total Examples | 1,229 |
| Mean Thinking Depth | ~1,667 words |
| Self-correction Present | 100% of traces |
| Verification Present | 100% of traces |
| Exploration Present | 100% of traces |
| Quality Gate Pass Rate | 100% |
Training Data Profile
- Category Mix: Math (1,016), Code (124), Science (45), Logic (44)
- Reasoning Depth: Premium traces average ~1,263 words of thinking vs ~281 for degenerate traces
- Drift Score Threshold: 1.463 cleanly separates premium from degenerate traces
- DDM AUC: 0.9705 | Sensitivity: 99.49% | False Positive Rate: ~5%
Intended Use
Designed for tasks requiring structured, multi-step reasoning:
- Mathematics
- Logic problems
- Code analysis
- Scientific problems
- Complex question answering
Limitations
- Single epoch training on 1,229 examples means the model retains most base Qwen 3.5 27B behavior; the fine-tune primarily shapes reasoning style rather than injecting new knowledge
- Language scope: DDM pipeline optimized for English; other languages reflect base model performance
- Edge cases: Extended thinking can occasionally loop on adversarial or highly ambiguous prompts
Citation
@misc{ornstein27b,
author = {DJLougen},
title = {Ornstein-27B: DDM-Curated Reasoning Fine-Tune of Qwen 3.5 27B},
year = {2026},
publisher = {Hugging Face},
url = {https://huggingface.co/DJLougen/Ornstein-27B}
}
Links
- Full-precision model: DJLougen/Ornstein-27B
- Base model: Qwen/Qwen3.5-27B | unsloth/Qwen3.5-27B