Create custom AI models trained on your data. Improve accuracy, reduce token usage, and build domain-specific intelligence that understands your business.
Upload examples in JSONL format
Model learns your patterns
Deploy and use via API
Get better results, faster responses, and lower costs with models trained on your data.
Models trained on your data understand your domain, terminology, and edge cases better.
Fine-tuned models often need fewer tokens to achieve the same result, reducing API costs.
Smaller, specialized models can respond faster than general-purpose alternatives.
Keep sensitive training data secure. Your fine-tuned model is exclusively yours.
Get reliable, predictable responses that match your brand voice and style.
Build models that understand legal, medical, financial, or technical terminology.
Create a JSONL file with example conversations or prompt-completion pairs. We recommend at least 100 examples for best results.
Upload your training file via API or dashboard. Choose your base model and configure hyperparameters like epochs and batch size.
Start the fine-tuning job. Training typically takes 30 minutes to a few hours depending on dataset size.
Once training completes, your model is immediately available via API with its own model ID.
from mythicdot import MythicDot client = MythicDot() # Upload training data training_file = client.files.create( file=open("training_data.jsonl", "rb"), purpose="fine-tune" ) # Start fine-tuning job job = client.fine_tuning.jobs.create( training_file=training_file.id, model="mythic-4-mini" ) # Check status print(f"Job ID: {job.id}") print(f"Status: {job.status}") # Once complete, use your model response = client.chat.completions.create( model="ft:mythic-4-mini:my-org:custom-model:abc123", messages=[{"role": "user", "content": "Hello!"}] )
Pay only for training and inference. No setup fees or minimum commitments.
| Model | Training (per 1K tokens) | Inference (per 1M tokens) |
|---|---|---|
| Mythic-4 Mini | $0.003 | $0.30 / $1.20 |
| Mythic-4 | $0.008 | $2.50 / $10.00 |
| Mythic-4 Ultra | $0.025 | $15.00 / $60.00 |
We recommend at least 100 examples for basic improvements. For best results, 500-1,000+ examples covering diverse scenarios will yield better performance.
Training time varies by dataset size and model. Small datasets (100-500 examples) typically complete in 30-60 minutes. Larger datasets may take several hours.
Yes, your training data is encrypted and used exclusively for your fine-tuning job. We do not use customer data to train base models.
Yes, you can delete fine-tuned models at any time from the dashboard or via API. Once deleted, the model is permanently removed.
Start fine-tuning today with our easy-to-use API and comprehensive documentation.