Global vs regional endpoints
Both clouds charge the Anthropic API price when requests can be served from any region: Bedrock through a global cross-region inference profile, Vertex AI through the global endpoint. Pinning traffic to one region costs more on every token type, including cache reads and writes. You may need a regional endpoint for data residency or for a quota that the global endpoint does not have.
| Model | Bedrock regional | Vertex AI regional |
|---|---|---|
| Claude Mythos 5.1 | $11 / $55+10% on the global rate | Not listed |
| Claude Fable 5.1 | $11 / $55+10% on the global rate | $11 / $55+10% on the global rate |
| Claude Fable 5 | $11 / $55+10% on the global rate | $11 / $55+10% on the global rate |
| Claude Opus 5.5 | $4.4 / $22+10% on the global rate | $4.4 / $22+10% on the global rate |
| Claude Opus 5 | $5.5 / $27.5+10% on the global rate | $5.5 / $27.5+10% on the global rate |
| Claude Opus 4.8 | $5.5 / $27.5+10% on the global rate | $5.5 / $27.5+10% on the global rate |
| Claude Sonnet 5.5 | $2.2 / $11+10% on the global rate | $2.2 / $11+10% on the global rate |
| Claude Sonnet 5 | $2.2 / $11+10% on the global rate | $2.2 / $11+10% on the global rate |
| Claude Sonnet 4.6 | $3.3 / $16.5+10% on the global rate | $3.3 / $16.5+10% on the global rate |
| Claude Haiku 5.5 | $0.11 / $0.55+10% on the global rate | $0.11 / $0.55+10% on the global rate |
| Claude Haiku 4.5 | $1.1 / $5.5+10% on the global rate | $1.1 / $5.5+10% on the global rate |
Prompt caching prices
Per 1M tokens on the Anthropic API. The global endpoint on each cloud charges the same where the table above says so.
| Model | Cache read | 5-min write | 1-hour write |
|---|---|---|---|
| Claude Mythos 5.1 | $0.25 | $12.5 | $20 |
| Claude Fable 5.1 | $0.25 | $12.5 | $20 |
| Claude Fable 5 | $1 | $12.5 | $20 |
| Claude Opus 5.5 | $0.2 | $5 | $8 |
| Claude Opus 5 | $0.5 | $6.25 | $10 |
| Claude Opus 4.8 | $0.5 | $6.25 | $10 |
| Claude Sonnet 5.5 | $0.1 | $2.5 | $4 |
| Claude Sonnet 5 | $0.2 | $2.5 | $4 |
| Claude Sonnet 4.6 | $0.3 | $3.75 | $6 |
| Claude Haiku 5.5 | $0.01 | $0.125 | $0.2 |
| Claude Haiku 4.5 | $0.1 | $1.25 | $2 |
Batch prices
Input / output per 1M tokens for asynchronous Batch jobs. This is where the platforms differ most.
| Model | Anthropic API | Bedrock | Vertex AI |
|---|---|---|---|
| Claude Mythos 5.1 | $5 / $25 | No Batch rate | Not listed |
| Claude Fable 5.1 | $5 / $25 | No Batch rate | $5 / $25 |
| Claude Fable 5 | $5 / $25 | No Batch rate | $5 / $25 |
| Claude Opus 5.5 | $2 / $10 | No Batch rate | No Batch rate |
| Claude Opus 5 | $2.5 / $12.5 | $2.5 / $12.5 | $2.5 / $12.5 |
| Claude Opus 4.8 | $2.5 / $12.5 | No Batch rate | $2.5 / $12.5 |
| Claude Sonnet 5.5 | $1 / $5 | No Batch rate | $1 / $5 |
| Claude Sonnet 5 | $1 / $5 | No Batch rate | $1 / $5 |
| Claude Sonnet 4.6 | $1.5 / $7.5 | $1.5 / $7.5 | $1.5 / $7.5 |
| Claude Haiku 5.5 | $0.05 / $0.25 | No Batch rate | No Batch rate |
| Claude Haiku 4.5 | $0.5 / $2.5 | $0.5 / $2.5 | $0.5 / $2.5 |
Choosing where to run Claude
- If your company already spends on AWS or Google Cloud, the same per-token price can come out of existing credits or committed spend.
- Use the global endpoint unless you need data to stay in one region; the regional uplift applies to every token.
- For large offline jobs, check the Batch table first. Without a Batch rate you pay the full standard price.
- Provisioned Throughput on Bedrock and Vertex AI is priced per model unit and is not compared here.
Sources
Amazon Bedrock: the AWS Price List file for Amazon Bedrock (also shown on the Bedrock pricing page), checked 2026-11-01. Google Vertex AI: partner model pricing, checked 2026-11-01. Anthropic API: Anthropic pricing.
Related pricing pages
Every Claude model and its sources are on the Anthropic API pricing page. To compare Claude with other providers, use the model pricing comparison or the prompt caching savings calculator.
Frequently asked questions
Is Claude cheaper on Amazon Bedrock or Google Vertex AI than on the Anthropic API?
No. On the global endpoint, all 21 listings compared here cost exactly the same as the Anthropic API for input, output and prompt caching. Claude Sonnet 5.5, for example, is $2 / $10 per 1M input / output tokens on all three. The price difference comes from regional endpoints, Batch availability and any discounts on your cloud account.
How much more do regional endpoints cost?
10% more on every token type, on both Bedrock (in-region inference without a cross-region profile) and Vertex AI (regional and US or EU multi-region endpoints). Claude Sonnet 5.5 goes from $2 / $10 to $2.2 / $11 per 1M tokens.
Does Amazon Bedrock offer Batch pricing for Claude?
Only for some models: Claude Opus 5 ($2.5 / $12.5 Batch), Claude Sonnet 4.6 ($1.5 / $7.5 Batch), Claude Haiku 4.5 ($0.5 / $2.5 Batch). AWS lists no Batch rate for the newer Claude models on Bedrock, while the Anthropic API Message Batches API halves the price for every model compared here.
Which Claude models are available on Vertex AI?
Claude Fable 5.1, Claude Fable 5, Claude Opus 5.5, Claude Opus 5, Claude Opus 4.8, Claude Sonnet 5.5, Claude Sonnet 5, Claude Sonnet 4.6, Claude Haiku 5.5, Claude Haiku 4.5. Claude Mythos 5.1 is listed on Amazon Bedrock only. Vertex AI publishes no usable Batch rate for Claude Opus 5.5 or Claude Haiku 5.5.
Is prompt caching priced the same on Bedrock and Vertex AI?
On the global endpoint, yes: cache reads, 5-minute cache writes and 1-hour cache writes match the Anthropic API for every listing that publishes them. Claude Sonnet 5.5 reads cached input at $0.1 per 1M tokens instead of $2. Regional endpoints add the same uplift to cache prices.
Where do these prices come from?
Bedrock rates come from the AWS Price List for Amazon Bedrock, Vertex AI rates from Google Cloud's partner model pricing page, and Anthropic API rates from Anthropic's pricing page. Each is a verified record in the AICostBudget pricing dataset with its check date shown in the tables.