Supervised fine-tuning (SFT)
Supervised fine-tuning (SFT) trains a model using labeled input-output pairs. The model learns from demonstration examples consisting of prompts and responses, refining its capabilities to align with specific tasks, instructions, or desired behaviors.
When to use SFT
Use SFT when you can specify what the right behavior looks like through labeled examples. SFT is ideal when:
-
You have high-quality input-output pairs that demonstrate the desired behavior
-
You want to teach the model a specific response format, tone, or style
-
Your task requires following domain-specific instructions or workflows
-
You need to adapt the model for multimodal tasks (text, image, video, or tool calling)
Supported models
SFT is available for the following Amazon Nova models:
-
Nova 1.0 (Micro, Lite, Pro)
-
Nova 2.0 (Lite)
When to use Nova 1.0 versus Nova 2.0
The Amazon Nova family of models offers multiple price-performance operating points to optimize between accuracy, speed, and cost.
Choose Nova 2.0 when you need the following:
-
Enhanced reasoning abilities with explicit reasoning mode support
-
Broader multilingual performance across additional languages
-
Improved performance on complex tasks including coding and tool use
-
Extended context handling with better accuracy and stability at longer context lengths
Note
The larger model is not always better. Consider the cost-performance tradeoff and your specific business requirements when selecting between Nova 1.0 and Nova 2.0 models.
