Nova Micro
The smallest, cheapest text-only Nova, built for ultra-high-throughput tasks and latency-sensitive routing.
At a glance
Released
Apr 22, 2026
Provider
Amazon
Context window
128,000 tokens
Max output
8,000 tokens
Modalities
Text → text
Knowledge cutoff
February 2026
API model ID
nova-micro-v1License
Proprietary (API-only via AWS Bedrock)
Pricing
Standard
Input$0.035 / 1M
Output$0.14 / 1M
Batch (50% off)
Input$0.0175 / 1M
Output$0.07 / 1M
A typical 1,500-token prompt with a 650-token answer costs ≈ $0.0001 on the standard tier. Run your own prompt through the calculator →
What's new vs the previous model
Compared withNova Micro (2024) (December 2024)
- Improved short-form accuracy and lower latency
- Cheaper per token than the prior generation
- Dropped any multimodal pretense; strictly text
Best use cases
Request routing
Classifies and dispatches traffic to larger models in a cascade.
Lightweight extraction
Pulls fields and entities from short text at massive scale.
Guardrails
Runs cheap pre- and post-filters around heavier pipelines.
Strengths & limitations
Strengths
- Rock-bottom cost
- Extremely fast
- Simple to deploy on Bedrock
Limitations
- Text only
- Shallow on multi-step reasoning
How to test it
- No-code: open the Amazon Bedrock console and run a prompt in the playground.
- API: create a key in the Amazon Bedrock console and call the API (example below).
aws bedrock-runtime converse \
--model-id nova-micro-v1 \
--messages '[{"role": "user", "content": [{"text": "Explain tokenization in one paragraph."}]}]'