🧠 CNOS Distillation Studio
Create custom models from larger ones
Ollama
Checking...
Trainer
Checking...
GPU
-
⚡ Fast Mode
Best defaults, minimal config
🔧 Advanced Mode
Full control over all settings
📚
Teacher Model
🏠 Local
Recommended
☁️ Cloud API
🔧 Custom
Select Ollama Model
llama3.2:3b (2GB) ⭐ Recommended
codellama:7b (3.8GB)
deepseek-r1:8b (5.2GB)
Ready
Cloud Provider
OpenAI (GPT-4)
Anthropic (Claude)
Together.ai
API Key
Custom Endpoint URL
🎓
Student Model
Select Size Template
Tiny
0.5B params
Small
1.5B params
Medium
3B params
Best
Custom
Your model
Local
🦙 Ollama
📦 GGUF File
Select Local Ollama Model
Loading models…
—
Select Local GGUF File
Scanning…
—
Drop
.gguf
files in
data/models/
to have them appear here.
↺ Refresh
Compression Level
4-bit
Balanced
3-bit
Recommended
2-bit
Smallest
📊
Training Data
Auto-Generate Training Data
Query teacher model for examples
Knowledge Focus
💬 Chat
💻 Coding
🧩 Reasoning
📖 General
Diversity Level
Medium
Samples per task
Variations per prompt
⚡
Optimization
Speed vs Quality
Balanced
▶
Advanced Training Parameters
Epochs
Batch Size
Learning Rate
Temperature
Alpha (Distillation)
🚀 Start Distillation
⏹️ Stop
📈
Training Progress
Idle
Ready to start
0%
-
Step
-
Epoch
-
Quality
-
Samples
Learning Progress (Quality vs Teacher)
0%
Starting
Learning
Expert
✅
Model Ready!
-
Size
-
Compression
-
Quality
-
Speed Gain
🧪 Try Model
🚀 Deploy to CNOS
⬇️ Download .cnosmodel
🦙 Export to Ollama
📦
Model Library
🔄 Refresh
Loading models...
🧪 Test Your Model
×
Teacher
llama3.2:3b
Ask a question to compare responses...
Your Model
distilled-3bit
Ask a question to compare responses...
Send