whichdevtool?
Menu
StacksDirectory Pricing changes Glossary
VPS and cloud servers App hosting and PaaS Managed databases Transactional email APIs Email marketing and newsletters Domain registrars CDN, DNS and edge security Monitoring, uptime and error tracking Auth and identity Frontend and full-stack frameworks Backend frameworks UI component libraries AI model APIs File and object storage Payments and billing Background jobs, queues and workflows ORMs and database toolkits

Cloud AI model platform · by Amazon Web Services

Amazon Bedrock

AWS service offering Claude, Amazon Nova, Llama, Mistral, DeepSeek, gpt-oss and other models on your AWS bill, with batch, flex, priority and provisioned modes.

Visit aws.amazon.com ↗ Alternatives Paid EU region
Amazon Bedrock homepage
§1

Overview

Amazon Bedrock is AWS's managed service for foundation models. One API (and one AWS bill) reaches models from Anthropic (Claude), Amazon (Nova), Meta (Llama), Mistral, DeepSeek, OpenAI's open-weight gpt-oss and others. Some Claude models that Anthropic has retired from its own API remain available on Bedrock.

Pricing is per token and depends on region and serving tier. Standard on-demand is the base; Batch and Flex are half price for work that can wait; Priority costs 75% more for preferential capacity; Provisioned Throughput reserves model capacity by the hour on one- or six-month terms. Cross-Region inference profiles (global, or geographic such as eu. and us.) spread traffic; the geographic and in-Region options cost 10% more than global.

There is no Bedrock free tier; usage is billed to your AWS account from the first token. The advantages are procurement and data location: existing AWS contracts and regional processing, including EU inference profiles.

§2

Pricing and plans

Plan Price What you get
On-demand (Standard) Per token e.g. Amazon Nova 2 Lite $0.30 in / $2.50 out per 1M
Batch 50% off on-demand Asynchronous jobs
Flex tier 50% off For workloads that can wait
Priority tier +75% Preferential compute allocation
Provisioned Throughput Hourly 1-month or 6-month commitment
Prompt caching Reduced cached-input rate Select models such as Claude and Nova

Read 2026-09-29. The Bedrock pricing page renders most per-model tables with JavaScript, so model prices were checked against third-party trackers that copy AWS's published price list: Amazon Nova 2 Lite at $0.30 in / $2.50 out per 1M on global or base on-demand, with cache reads $0.075 and batch at half price, and about $0.33 / $2.75 for US regional profiles (regional and geo profiles carry a 10% premium, per AWS and Anthropic docs). Prices differ by region. Anthropic's Claude models on Bedrock are priced by AWS, not Anthropic; Anthropic's docs note regional endpoints cost 10% more than global ones. No Bedrock free tier was found. fromPrice is left empty because the cheapest per-model rate could not be confirmed from the rendered page.

Source: aws.amazon.com/bedrock/pricing/ · prices as published, excluding tax.

§3

What it does

Multi-vendor catalogue
Claude, Nova, Llama, Mistral, DeepSeek, gpt-oss and more under one AWS API.
Service tiers
Standard, Priority (+75%) and Flex (50% off) on supported models.
Batch inference
50% discount versus on-demand for asynchronous jobs.
Provisioned Throughput
Reserved capacity billed hourly with 1- or 6-month commitments.
Cross-Region inference
Global, EU, US and other geographic inference profiles; geographic ones cost 10% more.
Intelligent Prompt Routing
Routes between models in one family (e.g. Nova Pro and Nova Lite) to cut cost by up to 30%, per AWS.
§4

Strengths and limitations

Strengths

  • Models from several makers on one AWS bill and contract.
  • EU and other geographic inference profiles for data-location needs.
  • Batch, Flex and Priority options for cost versus latency.
  • Keeps older Claude models available after Anthropic retires them from its own API.

Limitations

  • No free tier.
  • Per-model prices vary by region and endpoint type, and the pricing page is hard to read without JavaScript.
  • Geographic and regional endpoints cost 10% more than global.
  • Provisioned Throughput requires 1- or 6-month commitments.
§5

Who it suits

Good for

  • Companies already running on AWS
  • Teams needing regional processing, including in the EU
  • Workloads mixing Claude with cheaper open models

Look elsewhere if

  • Hobby projects that want a free allowance
  • Developers without an AWS account who just need one model's API
§6

Alternatives to Amazon Bedrock

Tool From Free option Stages
Mistral API (Mistral Studio) AI model API (model maker)
Usage-based Free tier
OpenAI API AI model API (model maker)
Usage-based Paid
Baseten Inference platform (Model APIs and dedicated deployments)
Usage-based Trial
OpenRouter Model router (unified API)
— Free tier
Claude API AI model API (model maker)
Usage-based Trial
Cerebras Inference Inference provider (open-weight models)
Usage-based Trial

All AI model APIs →

§7

Questions

Does Amazon Bedrock have a free tier?

No Bedrock free tier was listed on 2026-09-29; usage is billed to your AWS account.

Is Claude cheaper on Bedrock or on Anthropic's API?

AWS sets Bedrock prices. Anthropic's docs say Bedrock regional endpoints for Claude 4.5 and later cost 10% more than global endpoints; compare the region you need.

How much is batch inference on Bedrock?

Batch is 50% cheaper than on-demand for supported models.

Can Bedrock keep inference in the EU?

Yes. Bedrock offers EU cross-Region inference profiles (eu.) and EU regions, at a 10% premium over global profiles.

Sources read for this page