Price checkPer million tokens
On-Demand (pay per token)Usage-basedProvisioned ThroughputCustomEnterprise (Bedrock + AWS deal)Custom
See all 3 plans

All Amazon Bedrock Plans & Pricing

Plan Monthly Annual Best For
View all features by plan (compare side-by-side)

On-Demand (pay per token)

  • Claude Opus 4.6: $15/1M input, $75/1M output (AWS-hosted)
  • Claude Sonnet 4.6: $3/1M input, $15/1M output
  • Claude Haiku 4.5: $0.80/1M input, $4/1M output
  • Llama 3.3 70B: $0.72/1M input, $0.72/1M output
  • Titan Text Premier: $0.50/1M input, $1.50/1M output
  • 50+ models across Anthropic, Meta, Mistral, AI21, Cohere, Stability, Amazon

Provisioned Throughput

  • Commit to 1-month or 6-month term
  • Dedicated capacity, predictable latency
  • 30-50% cheaper than on-demand at steady-state load

Enterprise (Bedrock + AWS deal)

  • Volume commits
  • Custom models via Bedrock Marketplace
  • AWS enterprise support
Pricing Alerts

Track Amazon Bedrock pricing

Get an email when Amazon Bedrock's pricing changes — plus the weekly SaaS Price Watch: verified price changes and deals across 3,000+ products. One-click unsubscribe.

See Amazon Bedrock Plans
Compare Amazon Bedrock with alternativesAdjust seats, lock a tier, add up to 2 more products side-by-side. Shareable URL.
Quick Answer
Last verified:
High confidence

Amazon Bedrock offers usage-based pricing from $0.500–$75.00 per million tokens as of August 2026 and custom pricing for larger requirements. Plan: On-Demand (pay per token) (usage-based). Custom pricing is available on request. Usage cost depends on the selected model and generation volume.

Use the interactive pricing calculator to estimate your exact cost based on team size and requirements.

  • Free tier: No free tier available

Amazon Bedrock offers 3 pricing tiers: On-Demand (pay per token), Provisioned Throughput, Enterprise (Bedrock + AWS deal). Paid plans include On-Demand (pay per token) (usage-based). The Provisioned Throughput plan is high-steady-state inference workloads.

Compared to other llm api providers software, Amazon Bedrock is positioned at the budget-friendly price point.

  • 5 documented hidden costs beyond list price
  • G2 rating: 4.3 out of 5 from 71 reviews (as of Jul 2026)
Last verified:

How much does Amazon Bedrock cost?

Amazon Bedrock offers usage-based pricing from $0.500–$75.00 per million tokens, with custom pricing available for larger requirements. Plans include On-Demand (pay per token) (usage-based), Provisioned Throughput (custom pricing), Enterprise (Bedrock + AWS deal) (custom pricing).

Amazon Bedrock Pricing Overview

Amazon Bedrock offers usage-based pricing from $0.500–$75.00 per million tokens and custom pricing for larger requirements. The On-Demand (pay per token) plan is usage-based and is designed for aws-native deployments needing multi-model routing. The Provisioned Throughput plan requires contacting sales for a custom quote and is designed for high-steady-state inference workloads. The Enterprise (Bedrock + AWS deal) plan requires contacting sales for a custom quote and is designed for large aws customers negotiating edps.

There are at least 5 documented hidden costs beyond Amazon Bedrock's list price, including implementation, training, and add-on fees.

This pricing was last verified in August 3, 2026.

See Amazon Bedrock Plans

Amazon Bedrock pricing is fully consumption-based — there is no monthly seat fee on the On-Demand (pay per token) tier, so you pay only for tokens processed. Model costs span a wide range: Nova Micro starts at $0.035/1M input tokens and $0.14/1M output tokens, while the flagship Nova Premier runs $2.50/1M input and $12.50/1M output. Teams with predictable high-volume workloads can reserve capacity via Provisioned Throughput; large enterprise deals are custom-quoted under the Enterprise (Bedrock + AWS deal) tier.

How Amazon Bedrock Pricing Compares

Compare Amazon Bedrock pricing against top alternatives in LLM API Providers.

Groq $0-$3.0 per million tokens Compare → Together AI $0.03-$9.95 per million tokens / hour Compare → Fireworks AI $0-$9 per million tokens / hour Compare →

Usage-Based Rates

Per-unit pricing for Amazon Bedrock API usage.

On-Demand (pay per token)

Model Input Output Cached Per
claude-opus-4-6-bedrock 200K ctx $15.00 $75.00 — 1M tokens
claude-sonnet-4-6-bedrock 200K ctx $3.00 $15.00 — 1M tokens
claude-haiku-4-5-bedrock 200K ctx $0.800 $4.00 — 1M tokens
llama-3-3-70b-bedrock 131K ctx $0.720 $0.720 — 1M tokens
titan-text-premier 32K ctx $0.500 $1.50 — 1M tokens
  • Bedrock adds ~0-10% markup over direct provider pricing on some models
  • Prompt caching (preview): up to 90% reduction on cache-hit tokens
  • Provisioned Throughput: commit to 1+ month for discount (30-50% off on-demand)

Compare Amazon Bedrock vs Alternatives

Before committing to Amazon Bedrock, compare pricing with these 3 alternatives in the same category.

VSGroq
Free

Prototyping and evaluation

Full comparison
VSTogether AI
From $0.03 per million tokens / hour

Variable-volume API usage

Full comparison
VSFireworks AI
Free

Variable-volume API usage

Full comparison

All Amazon Bedrock alternatives & migration guides

What Companies Actually Pay for Amazon Bedrock

Median per-1M-token pricing across 5 models
Input $0.300/1M
Output $2.50/1M
Flagship models in this provider's catalog
Model Input /1M Output /1M Blended /1M
amazon/nova-premier-v1 $2.50 $12.50 —
amazon/nova-pro-v1 $0.800 $3.20 —
amazon/nova-2-lite-v1 $0.300 $2.50 —
amazon/nova-lite-v1 $0.060 $0.240 —
amazon/nova-micro-v1 $0.035 $0.140 —
Review scores
G2 4.3out of 5 (71)
Third-party review aggregates, as of Jul 2026
Top pricing complaints
Cost and Pricing ComplexityLearning Curve and ComplexityLimited Customization and FlexibilityModel and Access Issues
Source: OpenRouter API — medians aggregated from 5 models routed. Reflects router-surface pricing (may include modest markup vs direct provider rates).

How Amazon Bedrock Pricing Compares

Software Starting Price Top Price
Amazon Bedrock $0.07 per million tokens $75 per million tokens
Anyscale Free $4.9591 per million tokens
Baidu ERNIE API $0.1 per million tokens $10 per million tokens
Cerebras Inference API $0.1 per million tokens $6 per million tokens
Claude API $0.1 per million tokens $75 per million tokens
Cloudflare Workers AI Free $4.881 per million tokens

Detailed pricing comparisons:

Browse all LLM API Providers pricing →

5 Amazon Bedrock Hidden Costs Beyond the List Price

Beyond the listed price, Amazon Bedrock has at least 5 documented hidden costs that can significantly increase total cost of ownership.

Watch for 5 hidden costs
  • Knowledge Bases Vector Storage $345 per month
    high 1 source
    industry "Knowledge Bases Vector Storage: If using Amazon OpenSearch Serverless as the default vector store for Knowledge Bases, a minimum baseline cost of approximately $345 per month is incurred for 2 OpenSearch Compute Units (OCUs) at $0."
  • Bedrock Guardrails $0.15 per 1,000 text units, $0.00075 per image
    medium 1 source
    industry "These charges apply per evaluation, meaning if Guardrails run on both input and output, both are billed."
  • CloudWatch Logging $0.50 per GB for ingestion, $0.03 per GB for storage
    medium 1 source
    industry "CloudWatch Logs charges $0.50 per GB for ingestion and $0.03 per GB for storage in regions like us-east-1, which can add significant overhead."
  • Inactive Custom Models
    high 1 source
    industry "Unused custom models can continue to generate storage costs, potentially totaling thousands of dollars yearly for organizations with many inactive models."
  • Associated Infrastructure $780/month
    high 1 source
    industry "Associated Infrastructure: Beyond Bedrock itself, supporting AWS services like Lambda functions (e.g., $200/month), S3 storage (e.g., $150/month), API Gateway (e.g., $100/month), and VPC endpoints (e.g."
Tip

Ask your Amazon Bedrock sales rep about these costs upfront. Getting them in writing before signing can save you from surprise charges later.

Full hidden costs breakdown →

Intelligence sourced from 2 independent sources
industry G2 B2B reviews
Key claims include inline source attribution. Data verified against multiple independent sources. 6 source citations total.

Is this pricing incorrect? — we'll verify and update it.