Loading page
Amazon
Nova 2.0 Lite
Nova 2.0 Lite is a chat model from Amazon in the Nova family, positioned at the "Lite" tier below the family's larger variants.
Input from
$0.165 / 1M tokens
across 1 provider
API Pricing
Cheapest on Amazon AWS — 33% below avg| Provider | Input / 1M | Output / 1M |
|---|---|---|
| $0.165 | $1.38 | |
| $0.330 | $2.75 |
Prices updated daily. Last check: Sep 23, 2026
Nova 2.0 Lite pricing by provider
Input, output, and batch rates, plus alternatives, for one provider at a time.
Performance & Benchmarks
Source: Artificial Analysis →Intelligence
13.4 / 100
Coding
23.0 / 100
Math
94.3 / 100
Output Speed
185 t/s
Latency (TTFT)
7.9s
Reasoning & Knowledge
- MMLU-Pro81.8%
- GPQA Diamond81.1%
- Humanity's Last Exam11.6%
Coding
- LiveCodeBench71.1%
Math
- AIME 202594.3%
Agentic & Tool Use
- Terminal-Bench Hard16.7%
- Terminal-Bench v2.116.1%
- τ²-bench72.8%
Instruction & Long Context
- IFBench70.7%
- Long-Context Reasoning60.3%
Benchmarks measured Sep 2026. Scores are independent evaluations, not vendor-reported.
Model Details
General
- Creator
- Amazon
- Modalities
- Text
Capabilities
- Tool Calling
- No
- Open Source
- No
Strengths & Limitations
Strengths
- Measured output throughput of approximately 212 tokens per second (Artificial Analysis), which is fast for a general-purpose chat model
- "Lite" tier positioning within Amazon's Nova family targets cost-sensitive, high-volume text workloads rather than maximum-capability tasks
- Part of a versioned family (Nova 2.0), so sibling tiers are available if a workload needs to be moved up or down in capability
- Fast token streaming makes long-form generation complete quickly once the response begins
- Backed by Amazon, giving it a path to availability through Amazon's own cloud inference channels alongside third-party providers
- Independent third-party performance measurement is available rather than only vendor-published numbers
Limitations
- Measured time to first token of roughly 14.2 seconds is high, which is a poor fit for latency-sensitive interactive UIs
- We do not track a confirmed context window for this model, so long-document suitability must be verified against Amazon's documentation
- Modality support (image, audio, video input) is not confirmed in our data and should not be assumed
- Tool calling and structured output behavior are not tracked in our database for this entry
- As a "Lite" tier model, it is positioned below larger Nova variants for the most demanding reasoning and agentic tasks
Key Features
•Chat/text generation interface for conversational and instruction-following workloads
•Measured output speed of ~212 tokens per second (Artificial Analysis)
•"Lite" cost-and-speed tier within the Amazon Nova 2.0 family
•Streaming token output for progressive response rendering
•Amazon-developed model with availability through cloud inference providers
•Family versioning (Nova 2.0) allowing comparison against earlier and larger Nova entries
•Third-party benchmarked latency and throughput figures for capacity planning
About Nova 2.0 Lite
Nova 2.0 Lite is a chat-oriented large language model built by Amazon as part of its Nova model family. The "Lite" name places it in the lower-cost, higher-throughput tier of that family rather than among its larger siblings, which is the usual position for models intended to handle broad general-purpose conversational and text-generation workloads at volume rather than the hardest reasoning tasks.
Independent measurements collected by Artificial Analysis put Nova 2.0 Lite at roughly 212 output tokens per second, which is a fast generation rate relative to many general-purpose chat models. The same measurements record a time to first token of about 14.2 seconds, which is unusually high; long first-token latency of that magnitude typically indicates a substantial amount of pre-response processing before streaming begins. The practical consequence is a model that streams quickly once it starts but is not well matched to workloads where the very first token needs to appear immediately.
We do not track a confirmed context window, modality list, or tool-calling specification for Nova 2.0 Lite in our database, so those details should be checked against Amazon's own model documentation before committing to an integration. For provider availability and current rates, see the pricing table on this page.
Common Use Cases
Nova 2.0 Lite suits high-volume text workloads where per-request cost and sustained generation speed matter more than immediate first-token response: batch summarization, content drafting, classification and tagging over large document sets, customer-support reply generation, and back-office text transformation pipelines. The measured ~212 tokens per second output rate means long responses finish quickly once generation starts, which favors asynchronous or queued jobs. The high measured time to first token of about 14 seconds makes it a weaker choice for typing-speed chat interfaces, voice assistants, or autocomplete-style features where users notice startup delay. For tasks that need the deepest reasoning in Amazon's lineup, a larger Nova tier is the more natural comparison point.
Frequently Asked Questions
How much does Nova 2.0 Lite cost?
Pricing varies by provider and by pricing type — input tokens, output tokens, and any cached or batch rates are billed differently, and each host sets its own rates. Check the pricing table on this page for current per-provider figures rather than relying on a fixed number.
What is Nova 2.0 Lite best used for?
It is aimed at high-volume, cost-sensitive text generation: summarization, drafting, classification, and batch text processing. Its fast measured output rate of around 212 tokens per second helps on long generations, while its high measured time to first token makes it less suitable for real-time interactive chat.
How does Nova 2.0 Lite compare to larger Nova models?
The "Lite" label places it in the lower-cost tier of the Nova 2.0 family, below the larger variants. In practice that means it is the tier you reach for when throughput and unit cost drive the decision, and you step up to a bigger Nova model when a task needs more capability.
Is Nova 2.0 Lite fast?
It depends which part of latency you care about. Artificial Analysis measures roughly 212 output tokens per second, which is fast generation, but a time to first token of about 14.2 seconds, which is slow to start. Throughput-bound batch jobs benefit; latency-bound user-facing chat does not.
Does Nova 2.0 Lite support image input or tool calling?
Our database does not record confirmed modality or tool-calling details for this entry. Check Amazon's official model documentation and your chosen provider's API reference before designing around vision input or function calling.