Nova 2 Lite
Amazon Nova 2 Lite is a proprietary foundation model with 1M-token context, extended thinking, multimodal input, document and video understanding.
Model Overview
Capabilities, design details, and architectural traits
Nova 2 Lite - Reasoning foundation model
Nova 2 Lite is Amazon’s proprietary foundation model for production-scale agentic workflows, combining extended thinking with multimodal understanding and long-context processing.
Defining Characteristics
- 1 million-token context window: AWS documents Nova 2 Lite as supporting a 1M-token context, enabling very large prompts and multi-step agent loops.
- Extended thinking with three intensity levels: The model supports extended thinking and three-level intensity control to trade off speed and depth of reasoning.
- Multimodal input support: Official docs list text, image, video, and document inputs, with outputs centered on text generation.
- Document and video understanding: AWS highlights improved PDF, spreadsheet, table, chart, and multi-page document processing, plus stronger temporal video analysis.
Benchmark Performance
Independent evaluations · Artificial Analysis
Accuracy & Capability Details
Compare Models Side-by-Side
Evaluate specifications, pricing, and independent benchmark indices
| Model Details | |||
|---|---|---|---|
| General Info | |||
| Provider | Amazon | Anthropic | Anthropic |
| Release Date | October 29, 2025 | September 22, 2026 | September 28, 2026 |
| Knowledge Cutoff | - | - | Jun 2026 |
| Context & Limits | |||
| Context Window | 1M | 1M | 1M |
| Pricing (per 1M tokens) | |||
| Input Pricing | $0.33 Best Input Pricing | $4 | $2 |
| Output Pricing | $2.75 Best Output Pricing | $20 | $10 |
| Modalities | |||
| Inputs | textimagevideofile | textimagefile | textimagefile |
| Outputs | text | text | text |
| Benchmarks (0-100) | |||
| Intelligence Index | 11.8 | 57.6 Best Intelligence Index | 56.0 |
| Coding Index | - | - | - |
| Agentic Index | - | - | - |
Humanity's Last Exam
Extremely difficult logical reasoning and knowledge.
Long Context Reasoning
Logical reasoning over long context windows.
SciCode Benchmark
Scientific coding and mathematical modeling.
Independent evaluation data provided by Artificial Analysis. To view the latest benchmarks and full details, visit their official site.
Explore more from Amazon
Other models by Amazon
Top AI Models
Leading alternatives by intelligence score
