Amazon AI Guide: Titan & Nova Series Comparison

info

For detailed information about using models with APIpie, check out our Models Overview and Completions Guide.

Description

Amazon's Titan and Nova models represent AWS's proprietary family of foundation models, available through Amazon Bedrock. These models are designed to handle a variety of tasks including text generation, image generation, and embeddings, offering enterprise-grade reliability and performance through APIpie's routing system.

Model Families

Titan Models:

• Text Series: Ranging from lite to premier versions, optimized for different performance and resource needs
• Embedding Models: Specialized for text and image embeddings, ideal for search and recommendation systems
• Image Generation: Advanced models for creating and manipulating images

Nova Models:

• Text Processing: Specialized in handling extremely long contexts up to 300K tokens
• Performance Tiers: From micro to pro versions, balancing efficiency and capability
• Specialized Variants: Canvas and Reel versions for specific use cases

Key Features

• Extended Token Capacity: Models support context lengths from 4K to 300K tokens for various processing needs.
• Diverse Capabilities: Text generation, image generation, and embedding models available.
• Enterprise Focus: Built for production-grade applications with AWS's reliability.
• Optimized Performance: Models ranging from lightweight to high-performance versions.

---

Model Comparison

When choosing between Titan and Nova models, consider these key differences:

Titan Models:

• Best for: General-purpose text generation, image tasks, and embeddings
• Strengths:
  • More versatile with text, image, and embedding capabilities
  • Lower latency for standard tasks
  • Optimized for production workloads
• Use when:
  • You need image generation or embedding capabilities
  • Working with standard context lengths
  • Requiring fast response times

Nova Models:

• Best for: Long-form content and complex document processing
• Strengths:
  • Much larger context windows (up to 300K tokens)
  • Specialized for long-form content
  • Advanced text processing capabilities
• Use when:
  • Processing very long documents
  • Needing extensive context for responses
  • Working with complex, multi-part texts

---

Model List

Model List updates dynamically please see the Models Route for the up to date list of models

info

For detailed information about model capabilities and performance, visit the Amazon Titan Documentation.

Model Name	Max Tokens	Response Tokens	Providers	Type
titan-tg1-large	32,000	32,000	Bedrock	LLM
titan-text-lite-v1	4,000	4,000	Bedrock	LLM
titan-text-express-v1	8,000	8,000	Bedrock	LLM
titan-text-premier-v1	32,000	32,000	Bedrock	LLM
olympus-premier-v1	-	-	Bedrock	LLM
titan-embed-g1-text-02	8,192	8,192	Bedrock	Embedding
titan-embed-text-v1	8,192	8,192	Bedrock	Embedding
titan-embed-text-v2	-	-	Bedrock	Embedding
titan-image-generator-v1	-	-	Bedrock	Image
titan-image-generator-v2	-	-	Bedrock	Image
titan-embed-image-v1	-	-	Bedrock	Image
nova-pro-v1	300,000	5,120	Bedrock, OpenRouter	LLM
nova-lite-v1	300,000	5,120	Bedrock, OpenRouter	LLM
nova-micro-v1	128,000	5,120	Bedrock, OpenRouter	LLM
nova-canvas-v1	-	-	Bedrock	LLM
nova-reel-v1	-	-	Bedrock	LLM

---

Example API Call

Below is an example of how to use the Chat Completions API to interact with a Titan model.

curl -L -X POST 'https://apipie.ai/v1/chat/completions' \
-H 'Content-Type: application/json' \
-H 'Accept: application/json' \
-H 'Authorization: Bearer <YOUR_API_KEY>' \
--data-raw '{
  "provider": "bedrock",
  "model": "titan-text-premier-v1",
  "max_tokens": 150,
  "messages": [
    {
      "role": "user",
      "content": "Can you explain how photosynthesis works?"
    }
  ]
}'

---

Response Example

The expected response structure for a Titan model might look like this:

{
  "id": "chatcmpl-12345example12345",
  "object": "chat.completion",
  "created": 1729535643,
  "provider": "bedrock",
  "model": "titan-text-premier-v1",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Photosynthesis is the process by which plants convert sunlight into chemical energy. Here's how it works:\n\n1. **Light Absorption**: Plants capture sunlight using chlorophyll in their leaves\n\n2. **Water and CO2**: They take in water through roots and carbon dioxide through leaf pores\n\n3. **Chemical Reaction**: Using sunlight's energy, they convert H2O and CO2 into glucose and oxygen:\n   6CO2 + 6H2O + light → C6H12O6 + 6O2\n\nThis process produces food for the plant and releases oxygen as a byproduct."
      },
      "logprobs": null,
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 15,
    "completion_tokens": 125,
    "total_tokens": 140,
    "prompt_characters": 45,
    "response_characters": 520,
    "cost": 0.002250,
    "latency_ms": 3100
  },
  "system_fingerprint": "fp_123abc456def"
}

---

API Highlights

• Provider: Specify "bedrock" as the provider or leave blank for automatic selection.
• Model Selection:
  • Use Titan models for general text generation and embeddings
  • Use Nova models for extended context processing
  • See Models Guide for the complete list
• Max Tokens: Set according to model capacity (varies by model)
• Messages: Format your request following the message formatting guide

---

Applications and Integrations

Titan Applications:

• Text Generation:
  • Content creation and chat applications
  • Premier variant (32K context) for complex tasks
  • Express variant (8K context) for balanced performance
  • Lite variant (4K context) for efficient processing
• Image Generation:
  • Creating and editing images
  • Visual content generation
  • Image manipulation and enhancement
• Embeddings:
  • Semantic search implementation
  • Document similarity analysis
  • Content recommendation systems
  • Cross-modal applications with image embeddings

Nova Applications:

• Extended Context Processing:
  • Document analysis up to 300K tokens
  • Long-form content generation
  • Complex document summarization
• Specialized Processing:
  • Canvas variant for visual-heavy content
  • Reel variant for sequential content
  • Micro variant for efficient processing of medium-length content (128K tokens)

Integration Options:

• Try the models with LibreChat or OpenWebUI

---

Ethical Considerations

Amazon's models are designed with responsible AI principles. Users should implement appropriate safeguards and consider potential biases. For guidance, see AWS's Responsible AI Guidelines.

---

Licensing

Titan and Nova models are available through AWS Bedrock. Usage is subject to AWS Service Terms. For detailed information, consult the Amazon Bedrock documentation.

tip

Try out the Titan and Nova models in APIpie's various supported integrations.