Amazon Nova Model Menu, Speed Tiers & the Real Bill 2026 Guide
Amazon Nova is priced to win: Micro starts at $0.035 per million input, far below rivals. The bill grows with the model, speed tier and extras like web grounding at $30 per 1,000. Here is the math.
Typical token rate
$0.035-$12.50/1M
Nova Micro input to Nova Premier output; usage-based on AWS Bedrock
Hidden fees
Yes
web grounding at $30 per 1,000, Nova Act by the hour, Priority speed tier
Free tier
None
no free plan; pure pay-as-you-go, though Micro rates are near zero
Cost transparency
Medium
scores 4 of 6 on our transparency checklist
Amazon Nova true cost, model by model
High· Verified July 15, 2026Amazon Nova is pure pay-as-you-go on AWS Bedrock with no free tier as of July 15, 2026, and the model sets the rate. Nova Micro is $0.035 per million input and $0.14 output, Nova Pro is $0.80 and $3.20, and the flagship Nova Premier is $2.50 and $12.50. Batch takes roughly 50 percent off. The extras add up: web grounding is $30 per 1,000 requests, Nova Act bills $4.75 an agent-hour, and Nova Canvas prices per image. So the model, the speed tier and the extra tools decide the bill, not a subscription.
- Nova Micro in /1M$0.035
- Nova Micro out /1M$0.14
- Nova Lite in /1M$0.06
- Nova Pro in /1M$0.80
- Nova Premier in /1M$2.50
- Nova Act /hr$4.75
- Web grounding /1k$30
- Nova Canvas /image$0.04
Amazon Nova Micro starts at $0.035 per million input, far below the $7.99 median across the 20 llm tools we track. Nova's whole pitch is price, though the mid tiers trail the frontier.
Amazon Nova savings across a big model menu
Amazon Nova carries no seats and no coupons, and offers no student or nonprofit pricing in July 2026. None is needed, since the rates already sit among the lowest anywhere. Your savings come from picking the right model and the right speed tier, not from chasing a discount.
Route volume to Nova Micro or Lite, since they cost a fraction of Premier. Send delay-tolerant work through batch or the Flex tier for roughly half off Standard, and reserve Priority for genuinely latency-sensitive calls. Meter web grounding carefully at $30 per 1,000, and keep Nova Act agents from running in parallel unless the work needs it. At AWS scale, Bedrock committed-use pricing sits below the public card, which the tactics below work through, and the Amazon Nova alternatives page shows where rivals land.
Route to Micro and Lite
Nova Micro at $0.035 in and Lite at $0.06 handle most volume work at a fraction of Premier's $2.50. Reserving the flagship for hard reasoning is the single biggest saving on a Nova bill.
Batch and Flex halve the rate
Batch processing takes roughly 50 percent off across the family, and the Nova 2 Lite Flex tier runs around half of Standard for delay-tolerant work. For overnight or bulk jobs, that halving needs only the right tier.
Meter web grounding
At $30 per 1,000 requests, grounding can outweigh tokens on a search-heavy app. Grounding only the calls that truly need live results, and answering the rest from the model, keeps that line from running the bill.
Bedrock committed use
If Nova rides an existing AWS commitment, its usage folds into Bedrock committed-use and provisioned-throughput pricing, where rates move below the public card as part of your broader cloud agreement.
Avoid Priority by default
The Priority speed tier runs near 1.8 times Standard. Using it only where latency genuinely matters, and Standard or Flex everywhere else, quietly keeps the effective rate down across a high-volume workload.
Negotiating Amazon Nova through AWS
The published Nova rates are fixed below volume, and the levers are engineering: model choice, batch, Flex, disciplined grounding. Negotiation runs through your AWS relationship, since Nova usage counts toward Bedrock spend, where committed-use and provisioned-throughput pricing already apply.
Two moves carry the weight, and both rest on Nova already being cheap, so the play is scale and cloud commitment rather than a headline cut.
Fold Nova into a Bedrock commitment
- Target
- AWS committed use
- Argument
- Nova usage counts toward broader AWS and Bedrock spend. If you have or can negotiate a committed-use agreement, the effective rate moves below the public card as part of your cloud commitment, not as a standalone Nova deal.
Anchor on a frontier rival on quality
- Target
- Any quality-sensitive deal
- Argument
- Nova is cheap but the mid tiers trail frontier models like Claude. If your task needs frontier quality, price Claude or Gemini against Nova Premier, and decide whether the saving is worth the capability gap rather than assuming price wins.
Route batch work to the batch rate
- Target
- High-volume pipelines
- Argument
- Batch processing takes roughly 50 percent off the token rate across the portfolio. Anything that can wait, embeddings refreshes, backfills, nightly classification, belongs there. It is the one discount nobody has to approve, so claim it before asking AWS for anything else.
When an Amazon Nova commitment is worth timing
The Nova rate card ignores the calendar, so there is no timing angle on the models themselves. Timing applies only to a broader AWS commitment, which follows Amazon's enterprise cycle. A cloud agreement you can sign near a quarter close often prices better, and Nova usage simply rides that deal rather than being negotiated on its own. On its own, the spend is usually too small to move.
Jan
Feb
Mar
Q-END
Apr
May
Jun
Q-END
Jul
Aug
Sep
Q-END
Oct
Nov
Dec
Q-END
Pro tip: Bundle Nova into the bigger AWS conversation, not a standalone one. Because the token rates are already low, the leverage is in your total Bedrock and cloud commitment, where Nova's usage improves as part of the whole.
Amazon Nova costs: fixed and movable
Aim requests where AWS can move. The published Nova rates are fixed below volume; the room is in a Bedrock committed-use rate and your model and tier choices.
Usually negotiable
- Bedrock committed-use rateHIGH
- Provisioned throughput pricingHIGH
- Model and speed-tier selectionHIGH
- Grounding and Nova Act volume termsMEDIUM
- Payment terms via AWS billingLOW
Rarely negotiable
- The published per-model token rates
- The $30 per 1,000 web grounding rate below volume
- The Nova Act agent-hour price
- The AWS Bedrock-only deployment
Amazon Nova negotiation email generator
Nova is already cheap, so the note below works a different angle: folding usage into an AWS Bedrock commitment, or benchmarking frontier quality against the low rate. Complete the fields and the draft lists rival models at live catalog rates. State your monthly Nova volume, set a frontier or multi-cloud rate beside it, ask how Nova folds into your Bedrock commitment, and give a decision date.
fold Nova usage into a broader AWS spend agreement
Hi Amazon Nova team, I lead tooling decisions at [Your company], and we are evaluating Amazon Nova Team seats for a team of 10-50 people. As part of this evaluation we are also looking at Claude, which comes in at $3 per 1M input, and Google Gemini at $1.25 per 1M input. Can you help us understand the value difference at your current rates? We are ready to commit to an annual term. What is the best rate you can offer on annual billing, and can you cap the renewal price in the contract? We are aiming to sign before the end of this quarter, and budget sign-off is already in place. Could you share a proposal covering the per-seat or per-credit rate, the renewal terms, and any programs we qualify for? Best regards, [Your name] [Your company]
Send it Tuesday to Thursday, and follow up once after 3 business days.
Before you send
- Have your monthly Nova token volume and any existing AWS commitment ready, since both shape the rate.
- Send midweek, because a note arriving Tuesday through Thursday clears faster than a Monday or Friday one.
- Do not lead with your budget. Let the AWS rep frame the committed-use terms first, then push on them.
- Name a frontier or multi-cloud rival by rate. The generator inserts its current token price for you.
- Ask whether web grounding and Nova Act usage can fold into the committed rate, since those extras add up.
- Follow up once after a few business days, then read continued quiet as a read on your position.
Amazon Nova cost mistakes on Bedrock
Each of these comes from how a broad, cheap model menu meets a few expensive extras, and all are avoidable early.
Running everything on Premier. Nova Micro at $0.035 handles most volume at a fraction of Premier's $2.50 input.
Ignoring web grounding cost. At $30 per 1,000, it can outweigh tokens on a search-heavy app, so meter it.
Defaulting to the Priority tier. It runs near 1.8 times Standard, so reserve it for latency-sensitive calls.
Running Nova Act agents in parallel. Each instance bills separately at $4.75 an hour, which multiplies fast.
Skipping batch and Flex. Each takes roughly half off, so delay-tolerant work should never run at Standard.
Assuming preview rates are final. Several Nova 2 tiers are in preview, so their rate card can still change.
Amazon Nova rivals beyond AWS and mid-tier
Nova is cheap and AWS-bound, so the alternatives matter most when you need frontier quality or a different cloud. The three below are drawn from our verified catalog. You need not switch. What helps is a benchmarked rival, run on your own task, so a Bedrock commitment rests on a real quality-versus-price comparison rather than Nova's low headline rate alone.
Claude
API input, $15 output, frontier quality
$3/1M
The frontier benchmark, also available on Bedrock. Naming Claude forces the question of whether Nova's mid-tier saving is worth the capability gap on hard tasks.
Google Gemini
input rate, on Google Cloud
$1.25/1M
A frontier model on a different cloud. The card when AWS lock-in is the objection, or when you want frontier quality closer to Nova's price band.
Mistral Large
input, EU-hosted, multi-cloud
$2/1M
A capable model outside the AWS stack with EU hosting. The alternative when data residency or multi-cloud flexibility matters more than the last cent of token cost.
Script“Nova is cheap, but our task needs frontier quality and we are weighing clouds. Claude is $3 in, Gemini $1.25. What Bedrock committed rate keeps this workload on Nova?”
Is Amazon Nova worth it? A price-first read
Nova is the value play in the category, full stop. Nova Micro and Lite sit far below almost every rival, batch takes roughly half off, and even the flagship Premier stays competitive for demanding work. For high-volume, cost-sensitive workloads already on AWS, that pricing is hard to argue with, and the breadth of models, image, speech and agentic, is genuinely wide.
The costs to watch are the extras and the lock-in. Web grounding at $30 per 1,000, Nova Act by the agent-hour, and the Priority speed tier can each outrun the cheap token rate, and everything runs only through AWS Bedrock. The mid tiers also trail frontier models like Claude, so price is not the only axis. There is no free tier, though Micro's near-zero rate makes that almost moot.
So route volume to Micro and Lite, batch and use Flex, meter grounding and agents, and reserve Premier for what needs it. At scale, fold Nova into a Bedrock commitment. The full menu sits on the Amazon Nova pricing page, and Nova wins whenever price and AWS integration matter more than frontier quality.
Amazon Nova pricing and discount FAQ
What does Amazon Nova cost per million tokens?
+
It depends entirely on the model, since Nova is pure pay-as-you-go on AWS Bedrock with no free tier. Nova Micro is $0.035 per million input tokens and $0.14 output, while Nova Lite is $0.06 and $0.24. Nova Pro is $0.80 and $3.20, and the flagship Nova Premier is $2.50 and $12.50. Batch takes roughly 50 percent off. On top, web grounding is $30 per 1,000 requests, Nova Act bills $4.75 an agent-hour, and image and speech models price on their own scales. So the model and the extras set the bill.
Why is Amazon Nova so cheap?
+
Price is Nova's deliberate positioning. Amazon built the family to undercut rivals on token cost. Nova Micro at $0.035 per million input and Lite at $0.06 sit far below most competitors, and batch processing halves them again. The tradeoff is that the mid tiers trail frontier models like Claude on demanding reasoning and coding, and everything runs only through AWS Bedrock. So Nova is genuinely cheap, but the low price buys competitive mid-tier quality on one cloud, not frontier capability everywhere. For cost-sensitive, high-volume work, that is often the right trade.
Does Amazon Nova have a free tier?
+
No, there is no free plan. Nova is pure pay-as-you-go on AWS Bedrock, so you pay for every token from the first call. That said, the entry rates are so low, Nova Micro at $0.035 per million input, that light testing costs almost nothing in practice. If you are on AWS, you can prototype cheaply without a free tier. There is no student or nonprofit program either, because the pricing already sits near the floor, so the savings come from model choice rather than a discount.
What is the cheapest Amazon Nova model?
+
Nova Micro, at $0.035 per million input tokens and $0.14 output, is the cheapest, a text-only model built for high-volume, cost-sensitive work. Nova Lite is next at $0.06 and $0.24 and adds multimodal input. Batch processing takes roughly half off both. For most volume tasks, Micro or Lite is the right choice, with Nova Pro and Premier reserved for work that genuinely needs stronger reasoning. Sending each task to the smallest model that meets your quality bar is the largest single lever on a Nova bill.
How much does Amazon Nova web grounding cost?
+
Web grounding on Nova Premier and Nova 2 Pro costs $30 per 1,000 requests, which is well above the token rate and easy to underestimate. An application that grounds heavily in live search can see grounding, not tokens, become the largest line on the invoice. The fix is to ground only the calls that truly need real-time web results, letting the model or your own retrieval handle the rest. Metering grounding on purpose is one of the most effective ways to hold down a Nova bill on a search-heavy workload.
What is Nova Act and how is it billed?
+
Nova Act is Amazon's agentic web-automation model, billed at $4.75 per agent-hour rather than per token. Parallel agents are charged separately per instance, so running several concurrently multiplies the hourly cost quickly. Human-in-the-loop wait time is excluded from billing, which helps. Because it bills by elapsed time, Nova Act cost depends on how long your agents run and how many at once, not on tokens. Budget it as a variable, time-based line, and avoid spinning up parallel agents unless the task genuinely requires the concurrency.
Does Amazon Nova run outside AWS?
+
No. Every Nova model runs solely through AWS Bedrock, so a multi-cloud or on-prem team cannot deploy it. That lock-in is a real cost in flexibility. If your stack is on Google Cloud or Azure, or you need on-prem, Nova is off the table regardless of its low price. It also means Nova rides the Bedrock release cycle, and several newer Nova 2 tiers sit in preview where the rate card can still change. If cloud flexibility matters, weigh a multi-cloud rival against Nova's savings.
How do you keep an Amazon Nova bill low?
+
Route volume to Nova Micro or Lite and reserve Premier for hard tasks, since the rate spans $0.035 to $2.50 per million input. Send delay-tolerant work through batch or the Flex tier for roughly half off, and avoid the Priority tier unless latency matters. Meter web grounding at $30 per 1,000, and keep Nova Act agents from running in parallel needlessly. At AWS scale, fold Nova into a Bedrock committed-use agreement. Those choices target the model menu and the extras, which is where a Nova bill actually grows.
Is Amazon Nova cheaper than Claude or Gemini?
+
On raw token price, usually yes. Nova Micro at $0.035 and Lite at $0.06 per million input undercut Claude at $3 and Gemini at $1.25 by a wide margin. Even Nova Premier at $2.50 stays competitive. The difference is quality and reach: Claude and Gemini are frontier models, stronger on hard reasoning, and available on more clouds. So Nova wins for high-volume, cost-sensitive work where mid-tier quality suffices, while a frontier rival is worth the premium when the task genuinely needs it. Benchmark both on your workload before deciding.
Explore Amazon Nova
Every page on Amazon Nova in one place, you are on cost guide.
Snapshot, score and verdict
How to get API access, limits, SDKs and what it costs
Every tier and the entry price
You are here
Compared and ranked vs peers
Price and feature change history
Browse the full Large Language Models category
Sources & verification
| Source | What was checked | Last checked |
|---|---|---|
| Amazon Nova official pricing | Verified plan prices, renewal rates and credit allowances | July 15, 2026 |
| Amazon Nova website | Official vendor website | July 15, 2026 |
| Amazon Nova pricing on ComparEdge | Current prices for every plan, with the cost calculator | July 15, 2026 |
Every fact on this Amazon Nova pricing page is tied to a named source and a verification date. Freshness-sensitive figures trace to the sources above; verify against the vendor before relying on them.