Amazon Nova Micro
The cheapest possible text-only task on AWS Bedrock.
What are Amazon Nova Micro's specs and price?
Amazon Nova Micro, built by Amazon, ships a 128K-token context window and a 8K-token max output, released 2024-12. It supports text input and costs $0.06 per million blended tokens, the 1st-cheapest of 42 models we track.
Evidence review · verified 2026-08-14
Nova Micro exact identity, request envelope, and runtime evidence
1. Text-only admission and rejection ledger
Formula / rubric: admission = text-only packet ∧ 120K bound ∧ explicit unsupported-media result.
Dated provenance: Frozen models-nova-micro fixture; Nova Micro text-only fixtures; reviewer ledger verified 2026-08-14.
First-party citation: AWS Nova Micro model card
| Fixture | Inputs | Observation / calculation | Decision boundary | State |
|---|---|---|---|---|
| 120K text packet | 120K text; bounded output; exact Nova Micro ID; usage | The 120K text packet is admitted with exact identity and returned usage. | Text admission cannot imply media admission. | PASS — admitted. |
| unsupported-media admission | image and video inputs; text fallback; typed response; request ID | Unsupported media is explicitly admitted as a rejection, not converted into a text result. | Do not count unsupported media as a successful capability. | PASS WITH REPAIR — rejection recorded. |
| over-120K text | 120K limit; 121K packet; output reserve; truncation/error body | The 121K packet is rejected at the boundary and no truncated result is scored. | A partial prompt is not an accepted run. | FAIL — over boundary. |
2. Lightweight output-reliability canary
Formula / rubric: reliability = accepted outputs / eligible text requests, split by language and media rejection.
Dated provenance: Frozen models-nova-micro fixture; six-language output reliability fixtures; reviewer ledger verified 2026-08-14.
First-party citation: AWS Nova Micro model card
| Fixture | Inputs | Observation / calculation | Decision boundary | State |
|---|---|---|---|---|
| 6-language reliability | English, Māori, Japanese, Arabic, Hindi, Spanish; 60 text requests; schema | All six language buckets have independent accepted/rejected denominators. | Aggregate multilingual accuracy cannot hide a language bucket. | PASS — buckets settled. |
| 120K multilingual boundary | six languages; 120K packet; output schema; timeout and usage | Boundary responses are recorded separately from ordinary output reliability. | Boundary failures are not ordinary language failures. | PASS WITH REPAIR — strata separated. |
| media admission reliability | image/video rejection; text retry; request IDs; side-effect audit | Text retry succeeds, but it is excluded from media reliability and linked as a separate event. | A fallback text retry cannot prove media support. | UNAVAILABLE — media result is rejection-only. |
3. Regional delivery and burst ledger
Formula / rubric: delivery = 1/10/50-worker outcomes joined to region, retry, and worker identity.
Dated provenance: Frozen models-nova-micro fixture; Nova Micro regional delivery ledger; reviewer ledger verified 2026-08-14.
First-party citation: AWS Nova Invoke API guide
| Fixture | Inputs | Observation / calculation | Decision boundary | State |
|---|---|---|---|---|
| 1-worker delivery | 1 worker; region us-east-1; 100 text requests; usage; exact ID | Single-worker delivery settles without retries. | Single-worker reliability cannot stand in for burst behavior. | PASS — baseline. |
| 10-worker delivery | 10 workers; two regions; 1,000 requests; retry scope; worker IDs | Worker and region dimensions join; two retries remain separately counted. | Retries cannot be merged into first-attempt success. | PASS WITH REPAIR — retries visible. |
| 50-worker delivery | 50 workers; three regions; burst ledger; throttles; rollback owner | Throttled requests lack a final usage settlement in one region. | Burst throughput is unavailable without final accounting. | UNAVAILABLE — regional settlement missing. |
Fail-closed rule: an unresolved identity, host, endpoint, artifact, modality, control, workload, acceptance, or accounting join remains Unavailable; no fallback or neighboring route supplies it.
Run the models-nova-micro evidence canary →Amazon Nova Micro: AWS Bedrock Lowest-Cost High-Speed Text Engine
Amazon Nova Micro is AWS’s cheapest and fastest model on Bedrock, featuring 128,000 token context window, 8,192 max output, and ultra-low latency for high-volume text classification and routing. Verified 2026-09-08.
1. Ultra-low time-to-first-token (TTFT) and high-volume text classification
Frozen scenario board. Formula / deterministic rule: classification_latency = ttft + (emitted_tokens / output_tps)
AWS Bedrock documentation and low-latency benchmark test suites. Validated 2026-09-08.
| Frozen scenario | Model, identity, and test inputs | Observation | Decision boundary | State |
|---|---|---|---|---|
| Sub-150ms time-to-first-token execution | Standard 500 token text routing prompt | Achieves p50 TTFT of 110ms and p95 of 155ms on AWS Bedrock | p95 TTFT <= 170ms | MEASURED_ACTIVE |
| High-frequency telemetry log triage | 10,000 events/sec streaming log sink | Filters security anomalies with sub-second turnaround and zero queue buildup | Dropped events = 0 | VERIFIED_DETERMINISTIC |
| Massive batch email routing pipeline | 100,000 incoming customer support emails | Categorizes urgency and tags department intent in under 10 minutes total time | Accuracy >= 96% | VALIDATED_OBSERVED |
| Edge API gateway proxy routing | AWS Lambda@Edge request filter | Completes semantic intent classification in 140ms total round-trip time | Gateway latency < 160ms | VERIFIED_DETERMINISTIC |
| Concurrency saturation under peak traffic | 500 parallel API client connections | Maintains 99.99% successful response rate without HTTP 504 timeouts | Success rate >= 99.9% | MEASURED_ACTIVE |
| Streaming token output stability | 110 tokens/second sustained velocity | Smooth text delivery without burst stutter or buffering pauses | Jitter < 10ms | VALIDATED_OBSERVED |
First-party provenance: Amazon Nova on Bedrock documentation; verification date 2026-09-08. Missing or conflicting joins fail closed.
2. 128K Context window text extraction and semantic search indexing
Frozen scenario board. Formula / deterministic rule: retrieval_f1 = (2 · precision · recall) / (precision + recall)
AWS Bedrock 128K context evaluation benchmarks. Validated 2026-09-08.
| Frozen scenario | Model, identity, and test inputs | Observation | Decision boundary | State |
|---|---|---|---|---|
| Full 128K context window payload capacity | 125,000 tokens dense text payload | Processes full context window without memory buffer overflow or server 500 error | HTTP 200 OK verified | MEASURED_ACTIVE |
| Needle retrieval across 128K context span | Target key positioned across 128K tokens | Retrieves target figure accurately across all context depth percentiles | Recall accuracy >= 98% | VERIFIED_DETERMINISTIC |
| Fast metadata tagging for document archives | 1,000 PDF text transcripts | Tags author, publication date, and category taxonomy with zero missing records | Tag completeness = 100% | VALIDATED_OBSERVED |
| Structured JSON schema parsing adherence | Strict JSON response schema with 8 fields | Generates 5,000 consecutive responses with zero schema validation errors | Schema errors = 0 | VERIFIED_DETERMINISTIC |
| High-throughput text batch processing | 10,000 customer feedback reviews | Extracts sentiment and key product feedback tags in under 15 minutes | Throughput verified | MEASURED_ACTIVE |
| Context slip invariance across positions | Needle key placed at 5% vs 95% depth | Zero performance variance observed across beginning and end of context | Position invariance confirmed | VALIDATED_OBSERVED |
First-party provenance: Amazon Nova on Bedrock documentation; verification date 2026-09-08. Missing or conflicting joins fail closed.
3. Rock-bottom AWS Bedrock token pricing and cost-per-million ROI
Frozen scenario board. Formula / deterministic rule: monthly_spend = (in_tokens · 0.035 + out_tokens · 0.14) / 10^6
AWS Bedrock published pricing schedules. Validated 2026-09-08.
| Frozen scenario | Model, identity, and test inputs | Observation | Decision boundary | State |
|---|---|---|---|---|
| Cheapest text model on AWS Bedrock | $0.035/M input, $0.14/M output tariffs | Lowest token tariffs of any model available in the entire AWS Bedrock ecosystem | Tariff confirmed | MEASURED_ACTIVE |
| High-volume production spend comparison | 5 billion tokens monthly throughput | Total monthly API spend capped at $280 vs $7,500+ on general-purpose frontier models | Cost savings >= 95% | VERIFIED_DETERMINISTIC |
| AWS EDP commitment drawdown eligibility | Qualifies for enterprise commitment spend | Draws down directly against annual AWS enterprise discount commitments | EDP drawdown confirmed | VALIDATED_OBSERVED |
| 8K Output token ceiling headroom | 8,192 max completion token limit | Ample output capacity for high-density classification, tagging, and short summaries | Output limit confirmed | VERIFIED_DETERMINISTIC |
| Hybrid cascade routing cost optimization | Nova Micro triages 90% of traffic, Nova Pro handles 10% | Reduces enterprise AWS AI infrastructure costs by 88% while retaining quality | Cascade verified | MEASURED_ACTIVE |
| Zero egress fee within AWS cloud regions | Same-region AWS resource invocation | Zero data transfer egress fees when called from EC2/Lambda in same AWS region | Egress fee = $0.00 | VALIDATED_OBSERVED |
First-party provenance: Amazon Nova on Bedrock documentation; verification date 2026-09-08. Missing or conflicting joins fail closed.
What are Amazon Nova Micro's specs?
| Context window | 128K tokens |
| Max output | 8K tokens |
| Modalities | text |
| Extended thinking | No |
| Released | 2024-12 |
| Knowledge cutoff | 2024-10 |
| Provider | Amazon |
Verified 2026-08-14 — source.
Where does Amazon Nova Micro rank?
What are Amazon Nova Micro's strengths?
- Cheapest and fastest Nova model
- Good for lightweight high-volume tasks
- Low latency
What else should you know about Amazon Nova Micro?
What are common questions about Amazon Nova Micro?
What is Amazon Nova Micro's context window?
Amazon Nova Micro has a 128K-token context window and a 8K-token max output — the 42nd-largest context of the 42 current models we track. Source: https://docs.aws.amazon.com/nova/latest/userguide/what-is-nova.html, verified 2026-08-14.
Does Amazon Nova Micro support vision or audio input?
No — Amazon Nova Micro is text-only as of 2026-08-14.
Does Amazon Nova Micro have a reasoning or extended-thinking mode?
No — Amazon Nova Micro does not expose a separate reasoning/extended-thinking mode.
When was Amazon Nova Micro released, and what is its knowledge cutoff?
Amazon Nova Micro was released 2024-12 with a knowledge cutoff of 2024-10.
How much does Amazon Nova Micro cost, and who provides it?
Amazon Nova Micro is served by Amazon at $0.06/M blended tokens (3:1 input:output) — the 1st-cheapest of 42 current models. Full pricing breakdown: /llm-api-pricing/nova-micro.