Amazon · Amazon Nova · Mid
Amazon Nova Pro pricing and limits
Amazon's multimodal mid-tier model on Bedrock. Priced well under frontier models, with Bedrock prompt caching and batch inference for further savings.
| Line | Tokens | Rate / 1M | Cost |
|---|---|---|---|
| Input, uncached | 5,000 | $0.80 | $0.00400 |
| Input, cache read | 10,000 | $0.20 | $0.00200 |
| Output | 800 | $3.20 | $0.00256 |
- Input / 1M
- $0.80
- Output / 1M
- $3.20
- Cached / 1M
- $0.20
- Cache write / 1M
- n/a
- Blended 3:1
- $1.40
- Context
- 300K
What drives Amazon Nova Pro cost
Bedrock prompt caching bills cache reads at a quarter of the input rate for Nova. Long system prompts and document context benefit, provided they sit in a stable prefix ahead of each request's variable content.
Running through Bedrock adds AWS-side considerations: cross-region inference profiles, provisioned throughput commitments and per-region pricing. PALM SOUQ meters at list on-demand rates per tenant so you can compare against your AWS invoice.
Cost of a typical call
Take a document Q&A turn with 15,000 input tokens, 10,000 of them served from cache, and 800 output tokens. At Amazon Nova Pro list prices ($0.80 input, $0.20 cached input, $3.20 output per 1M) one call costs $0.00856, against $0.0146 with no cache hits. At 200,000 calls a month that is $1,712, or $2,912 uncached.
Blended at a 3:1 input to output ratio, Amazon Nova Pro costs $1.40 per 1M tokens. Routing the same call to the fallback in the policy below, Amazon Nova Lite, costs $0.00064, which is what a tenant pays you in COGS once they cross the hard cap and PALM SOUQ downgrades instead of blocking.
Cost comparison
Amazon Nova Pro against three alternatives
The same document Q&A turn priced on Amazon Nova Lite, Claude Haiku 4.5, Gemini 2.5 Flash. Move the sliders to match your traffic; models without a cache-read rate bill every input token in full.
Benchmark quality per task before switching. A cheaper model that needs more steps or longer outputs can cost more per finished task than its rate suggests, which is why PALM SOUQ reports cost per feature, not only per token.
Amazon Nova Pro this guide
$0.00853 per call
$2,133 / mo
baseline
Amazon Nova Lite
$0.00064 per call
$160 / mo
-92% vs Amazon Nova Pro
Claude Haiku 4.5
$0.00996 per call
$2,489 / mo
+17% vs Amazon Nova Pro
Gemini 2.5 Flash
$0.00424 per call
$1,060 / mo
-50% vs Amazon Nova Pro
List prices from the PALM SOUQ LLM Price Index, computed per call as uncached input + cache reads + output. Reasoning tokens count as output on reasoning models.
Rate limits by usage tier
Bedrock quotas are set per account and per region as requests and tokens per minute for each model, adjustable through AWS Service Quotas. Cross-region inference can raise effective throughput.
Exact limits vary by account tier and change over time, so we do not reprint them here. Check the limits page for your own account in the Amazon documentation at docs.aws.amazon.com/bedrock. PALM SOUQ enforces your own per-tenant limits in front of the provider's, so a single customer cannot consume the shared allowance that every other tenant depends on.
Circuit-breaker best practices
Per-tenant breakers are especially useful on Bedrock because AWS quotas are shared by every application in the account; one runaway tenant can throttle unrelated workloads.
Use max step caps and token velocity ceilings together for Nova agents built with Bedrock Agents or your own orchestration.
When a breaker trips, the gateway answers in under 50 ms with HTTP 429 and X-PalmSouq-Reason: circuit_breaker_triggered, before the request reaches Amazon. Your agent framework sees a normal rate-limit response and stops, and the audit log records which heuristic fired and for which tenant.
# palmsouq.policy.yaml
policy: nova-pro-business
match:
model: nova-pro
tier: business # from X-PalmSouq-Tier
budgets:
monthly_usd: 200 # per tenant
soft_cap_pct: 80 # webhook to your app
on_hard_cap: fallback # or: block (HTTP 429)
fallback:
- nova-lite
breakers:
tool_thrashing: { identical_calls: 3, window_steps: 8 }
token_velocity: { max_tokens_per_s: 50000, window_s: 10 }
semantic_echo: { repeats: 3 }
max_steps: 15
cache:
meter_reads: true- budgetsPer-tenant monthly dollar cap with an 80% soft-cap webhook, keyed by X-PalmSouq-Tenant.
- fallbackAt the hard cap, requests reroute to Amazon Nova Lite ($0.10 blended per 1M) instead of failing.
- breakersAll four loop heuristics on, evaluated in the request path with under 5 ms added latency.
Where it fits
AWS-native teams that want to keep inference inside their cloud account, and multimodal document workloads at mid-tier prices.
Provider documentation: docs.aws.amazon.com/bedrock
Amazon Nova Pro pricing questions
How much does Amazon Nova Pro cost per million tokens?
At list price Amazon Nova Pro costs $0.80 per 1M input tokens and $3.20 per 1M output tokens, with cached input at $0.20 per 1M. Prices were last reviewed in the PALM SOUQ LLM Price Index and are for the standard tier and shortest context band.
Does Amazon Nova Pro support prompt caching?
Yes. Cached input is billed at $0.20 per 1M tokens against $0.80 for uncached input. PALM SOUQ records cached tokens per tenant from the provider's usage data, so cache savings show up in each customer's COGS.
What are the Amazon Nova Pro rate limits?
Bedrock quotas are set per account and per region as requests and tokens per minute for each model, adjustable through AWS Service Quotas. Cross-region inference can raise effective throughput. For your account's exact numbers, see the Amazon documentation at docs.aws.amazon.com/bedrock.
How do I stop runaway Amazon Nova Pro agent loops?
Per-tenant breakers are especially useful on Bedrock because AWS quotas are shared by every application in the account; one runaway tenant can throttle unrelated workloads. The sample policy on this page enables all four PALM SOUQ heuristics with a max step cap of 15 and falls back to Amazon Nova Lite at the hard cap.
Is there a batch discount for Amazon Nova Pro?
50% off on-demand rates for supported models through Bedrock batch inference. Batch traffic suits evaluations, backfills and nightly enrichment jobs that do not need a real-time answer.
Govern Amazon Nova Pro spend per customer.
Point your Amazon client at gateway.palmsouqltd.com, pass X-PalmSouq-Tenant, and every Amazon Nova Pro token is attributed, capped and protected by circuit breakers.
14-day free production trial · 100k requests included · no card required