Exploring Amazon Nova LLM Series: A Full Breakdown of Prices and Features
Description
Amazon introduced the Amazon Nova series of large language models (LLMs) during the AWS re:Invent conference. The series includes three versions: Micro, Lite, and Pro, directly competing with Google Gemini. This article provides a detailed analysis of Nova’s features, pricing, and comparisons with other major models in the market.
Overview and Highlights of the Nova Series
Amazon Nova debuted with impressive multimodal capabilities, supporting text, image, and video inputs (audio support is not yet available). Here are the core features of each model:
- Nova Micro: Affordable, suitable for text-based tasks.
- Nova Lite: Supports images and documents, ideal for medium workloads.
- Nova Pro: The most advanced multimodal capabilities, designed for high-end applications.
The Premier model in the Nova series is still under training and is expected to launch in 2025, offering over 2 million tokens for context processing.
Pricing and Feature Comparison
Amazon Nova offers competitive pricing. Here’s a comparison with other mainstream models:
Entry-Level Models
| Provider | Model | Per Million Input Tokens (USD) | Per Million Output Tokens (USD) |
|---|---|---|---|
| OpenAI | GPT-4o Mini | 0.15 | 0.6 |
| Gemini 1.5 Flash-8B | 0.0375 | 0.15 | |
| Gemini 1.5 Flash | 0.075 | 0.3 | |
| Amazon | Nova Micro | 0.035 | 0.15 |
| Amazon | Nova Lite | 0.06 | 0.24 |
| Anthropic | Claude 3 Haiku | 0.25 | 1.25 |
| Anthropic | Claude 3.5 Haiku | 1 | 5 |
High-End Models
| Provider | Model | Per Million Input Tokens (USD) | Per Million Output Tokens (USD) |
|---|---|---|---|
| OpenAI | GPT-4o | 2.5 | 10 |
| OpenAI | GPT-o1-mini | 3 | 12 |
| OpenAI | GPT-o1-preview | 15 | 60 |
| Gemini 1.5 Pro | 1.25 | 5 | |
| Anthropic | Claude 3.5 Sonnet | 3 | 15 |
| Anthropic | Claude 3 Opus | 15 | 75 |
| Amazon | Nova Pro | 0.80 | 3.2 |
Nova Pro’s pricing is slightly lower than Claude 3.5 Haiku, demonstrating a competitive edge in the high-end market.
For any incorrect pricing information, please contact our support for corrections.
Testing Nova’s Multimodal Capabilities
Nova Lite and Nova Pro support image and video processing. Below are test scenarios:
Image Description Generation Inputting an image from Discovery, Nova Pro generated a detailed description, covering scene, lighting, and object behavior. Cost: approximately $0.00242.
Video Processing Nova Pro analyzed sequences in a video but could not process audio content.
PDF Document Handling Nova Pro successfully converted complex PDFs into Markdown format. However, improvements are needed for handling tables and charts.
Conclusion: Amazon Nova’s Market Positioning
Strengths
- Competitive Pricing: Nova Micro is currently the cheapest model in the market.
- Multimodal Capabilities: Supports image and video inputs, broadening application scenarios.
- High Cost-Performance Ratio: Nova Pro offers advanced multimodal capabilities at a relatively low price.
Weaknesses
- Complex AWS Setup: High API access complexity might deter new users.
- Context Limitations: With 300,000 tokens, Nova lags behind Google Gemini’s 2 million tokens.
Future Outlook
Amazon plans to launch a new Nova model in 2025, featuring “cross-modal conversion” and voice input support. This could position the Nova series as a leader in multimodal AI.
Remarks
Overall, the Amazon Nova series brings fresh competition to the LLM market, excelling in pricing and multimodal capabilities. By competing directly with Google Gemini and enhancing its support for images and videos, Nova provides users with more choices. However, compared to Gemini’s ease of API access via direct endpoints, AWS’s higher complexity still requires improvement.
Despite these challenges, Nova demonstrates Amazon’s strong technical capabilities and determination to challenge top-tier model providers. If user-friendliness improves and API barriers are addressed, Nova could secure a significant place in the LLM market, potentially driving competitors to rethink their pricing strategies. For users, this creates a win-win situation.
