Amazon Nova 2 Lite

amazonamazon/nova-2-lite

API guide

Chat completion

Call Nova 2 Lite through TheRouter's OpenAI-compatible surface. TheRouter handles the Bedrock InvokeModel / Converse translation upstream — your client code stays OpenAI-shaped.

cURL

curl https://api.therouter.ai/v1/chat/completions \
  -H "Authorization: Bearer $THEROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "amazon/nova-2-lite",
    "messages": [{"role": "user", "content": "Summarise this 200-page PDF in 5 bullets."}]
  }'

Streaming

Stream tokens for chat UIs. With extended thinking enabled, time-to-first-token grows with the chosen budget; streaming hides the latency once generation starts.

cURL

curl https://api.therouter.ai/v1/chat/completions \
  -H "Authorization: Bearer $THEROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "amazon/nova-2-lite",
    "stream": true,
    "messages": [{"role": "user", "content": "Explain DSA attention in 200 words."}]
  }'

Tool use

Nova 2 Lite supports function calling through the standard tools/tool_choice fields. Note: web grounding and the code interpreter are managed by Bedrock as runtime tools, separate from your declared functions.

cURL

curl https://api.therouter.ai/v1/chat/completions \
  -H "Authorization: Bearer $THEROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "amazon/nova-2-lite",
    "messages": [{"role": "user", "content": "Look up the AWS region for Tokyo."}],
    "tools": [{
      "type": "function",
      "function": {
        "name": "lookup_region",
        "parameters": {"type": "object", "properties": {"city": {"type": "string"}}}
      }
    }]
  }'

Structured JSON

Use response_format to constrain output to JSON. Useful for downstream parsers; pair with a JSON schema in the system prompt for stricter shape control.

cURL

curl https://api.therouter.ai/v1/chat/completions \
  -H "Authorization: Bearer $THEROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "amazon/nova-2-lite",
    "response_format": {"type": "json_object"},
    "messages": [
      {"role": "system", "content": "Return JSON: {\"intent\":string,\"entities\":string[]}"},
      {"role": "user", "content": "Book a flight from Tokyo to Sydney next Friday."}
    ]
  }'

Vision input

Nova 2 Lite is natively multimodal — pass images alongside text in the OpenAI-compatible content array. Per AWS, document and short video inputs are also supported.

cURL

curl https://api.therouter.ai/v1/chat/completions \
  -H "Authorization: Bearer $THEROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "amazon/nova-2-lite",
    "messages": [{
      "role": "user",
      "content": [
        {"type": "text", "text": "Extract the invoice total and due date."},
        {"type": "image_url", "image_url": {"url": "https://example.com/invoice.png"}}
      ]
    }]
  }'

Extended thinking

Off by default. Enable via the reasoning field and pick a budget — low / medium / high — to trade speed for depth. AWS recommends starting off, then stepping up only on tasks that need it.

cURL

curl https://api.therouter.ai/v1/chat/completions \
  -H "Authorization: Bearer $THEROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "amazon/nova-2-lite",
    "reasoning": {"effort": "medium"},
    "messages": [{"role": "user", "content": "Plan a multi-step deployment migration."}]
  }'

Fact ledger — every claim on this page traces here

source	URL	retrieved
Release window	aws.amazon.com ↗	2026-05-22	verified
Bedrock model id	aws.amazon.com ↗	2026-05-22	verified
Training cutoff	—	—	unknown
Extended thinking budget levels	aws.amazon.com ↗	2026-05-22	verified
Built-in tools	aws.amazon.com ↗	2026-05-22	verified
Customization	aws.amazon.com ↗	2026-05-22	verified
Bedrock tiers	aws.amazon.com ↗	2026-05-22	to verify
License	—	—	verified
MMLU-Pro	medium.com ↗	2026-05-22	single source
GPQA Diamond	medium.com ↗	2026-05-22	single source
Artificial Analysis Intelligence Index	medium.com ↗	2026-05-22	single source
Long-context benchmarks (LongCodeBench 1M, OpenAI-MRCR)	assets.amazon.science ↗	2026-05-22	unknown
AWS introduces Nova 2 Lite — 1M context, extended thinking, web grounding built in	aws.amazon.com/blogs/aws ↗	2026-05-22	verified
Amazon publishes Nova 2 technical report	assets.amazon.science ↗	2026-05-22	verified
What is the context window of Nova 2 Lite?	aws.amazon.com ↗	2026-05-22	to verify
Does Nova 2 Lite support extended reasoning?	aws.amazon.com ↗	2026-05-22	to verify
What tools come built in?	aws.amazon.com ↗	2026-05-22	to verify
Can I fine-tune Nova 2 Lite?	aws.amazon.com ↗	2026-05-22	to verify
How does Nova 2 Lite pricing compare across Bedrock tiers?	aws.amazon.com ↗	2026-05-22	to verify