Claude Sonnet 5: EU IDs, Price, Bedrock 2026

TL;DR

Claude Sonnet 5 in 2026 is two IDs: current claude-sonnet-5-5 and legacy claude-sonnet-5. Both cost 2 / 10 USD per million tokens. In the EU, pin eu.anthropic.claude-sonnet-5-5 on Bedrock or the Europe multi-region on Google Cloud. Foundry and the Claude API have no EU inference. New workloads start on Sonnet 5.5.

What is Claude Sonnet 5 in 2026?

Claude Sonnet 5 is the production model for speed and intelligence. A 2026 search hits two live IDs: Claude Sonnet 5.5 (current, 28 September 2026) and Claude Sonnet 5 (legacy, 30 June 2026). Both cost 2 / 10 USD. Both have a 1 million token context. New pipelines pin Sonnet 5.5.

The models overview we read on 4 October 2026 names four current models: Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 4.5. Claude Sonnet 5 sits on the legacy list. Anthropic’s Sonnet 5 page tells you to migrate to 5.5. The list price does not change. The gap is knowledge cutoff, thinking controls, and the IDs you pin on Bedrock and Google Cloud.

Feature	Claude Sonnet 5.5	Claude Sonnet 5
Status	Active (latest)	Active (legacy)
Released	28 September 2026	30 June 2026
Retirement (Anthropic)	Not sooner than 28 September 2027	Not sooner than 30 June 2027
Claude API ID	claude-sonnet-5-5	claude-sonnet-5
List price in / out	2 / 10 USD per MTok	2 / 10 USD per MTok
Context / max output	1M / 128K	1M / 128K
Knowledge cutoff	June 2026	January 2026
Default effort	high	high
Turn thinking off	between_tools, not disabled	"thinking": {"type": "disabled"} works

One million tokens is about 555,000 words on the tokenizer introduced with Claude Opus 4.7. That tokenizer produces about 30 percent more tokens than the older one. A budget written on a Sonnet 4.x bill sits too low. The cost-per-task split against Opus 5.5 is in Claude Sonnet vs Opus for company workloads. Full token tables sit in Claude API pricing in the EU.

Five breaking changes hit code still running on Sonnet 5. The Sonnet 5.5 page lists them: turn off up-front thinking with between_tools, not disabled. Forced tool use returns an error. Thinking blocks are tied to the model and the conversation. On the Claude API and Google Cloud, the older computer_20251124 computer-use tool is rejected. The advisor tool rejects Claude Sonnet 5 as an advisor. A non-default temperature, top_p or top_k returns a 400 on both Sonnet 5 IDs.

What does Claude Sonnet 5 cost on the API, Bedrock and Google Cloud?

Claude Sonnet 5 and Claude Sonnet 5.5 cost 2 USD input and 10 USD output per million tokens on the Claude API. Batch halves that to 1 / 5 USD. Cache reads cost 0.20 USD. Regional and multi-region endpoints add 10 percent from Sonnet 4.5 on. Google lists 2.20 / 11 USD. No euro list price exists.

The numbers come from 4 October 2026, from Anthropic’s pricing page and Google Cloud’s partner table. Footnote 3 on the Anthropic page states that the 2 / 10 USD Sonnet 5 rate, announced as introductory pricing through 31 August 2026, is now the standard price. The increase to 3 / 15 USD planned for 1 September 2026 will not occur.

Route	Input / output	5-min cache write	Cache read	Batch in / out
Claude API global	2.00 / 10.00	2.50	0.20	1.00 / 5.00
Google Cloud, first partner table	2.00 / 10.00	2.50	0.20	1.00 / 5.00
Google Cloud, later tables (10% row)	2.20 / 11.00	2.75	0.22	1.10 / 5.50
Bedrock EU geo, 10% rule	2.20 / 11.00	2.75	0.22	50% of the EU row

USD per million tokens. The same table applies to Sonnet 5 and Sonnet 5.5. Foundry and Claude Platform on AWS bill in Claude Consumption Units at 0.01 USD per CCU. One hundred CCU equals 1.00 USD of list fees. That is a billing unit, not a discount. Anthropic applies the 10 percent rule from Sonnet 4.5, Haiku 4.5 and Opus 4.5 on Bedrock regional endpoints and Google multi-regions.

A worked example, our assumption, not a vendor quote: 10,000 support replies at 3,000 input and 400 output tokens cost 100 USD on Sonnet 5.5 globally. The same volume on the 10 percent row costs 110 USD. Batch on that row costs 55 USD. Prompt caching pays for itself after one 5-minute cache read, or after two 1-hour reads.

US-only inference on the Claude API (inference_geo: "us") costs 1.1x. That is not an EU product. The data residency page lists only global and us. Teams that need processing in EU member states pay the 10 percent on Bedrock or Google Cloud, not on the Claude API.

Which model IDs apply in the EU?

For Claude Sonnet 5.5, pin eu.anthropic.claude-sonnet-5-5 on Bedrock and claude-sonnet-5-5 with region="eu" on Google Cloud. The legacy ID eu.anthropic.claude-sonnet-5 still works. The bare Bedrock ID anthropic.claude-sonnet-5-5 is the card ID, not the modelId you send to bedrock-runtime. Foundry and the Claude API use claude-sonnet-5-5 with no EU prefix.

The IDs sit on the Sonnet 5.5 and Sonnet 5 pages and on the Bedrock cards. AWS samples still call global.anthropic.claude-sonnet-5-5 from us-east-1. Copy that into a Frankfurt account and the data leaves the EU.

Platform	Sonnet 5.5	Sonnet 5 (legacy)	EU processing
Claude API	claude-sonnet-5-5	claude-sonnet-5	No. Only global or us
Amazon Bedrock, EU geo	eu.anthropic.claude-sonnet-5-5	eu.anthropic.claude-sonnet-5	Yes. Data stays in EU regions
Amazon Bedrock, global	global.anthropic.claude-sonnet-5-5	global.anthropic.claude-sonnet-5	No
Google Cloud	claude-sonnet-5-5	claude-sonnet-5	Yes, on the Europe multi-region
Microsoft Foundry	claude-sonnet-5-5	claude-sonnet-5	No EU data zone

On Bedrock, Sonnet 5 also publishes AU and India geo IDs. Sonnet 5.5 lists US geo, EU geo and Global. In-region on bedrock-runtime is N/A for Sonnet 5.5. The Sonnet 5 mantle table names eu-north-1 (Stockholm) and eu-west-1 (Ireland) in Europe. The Sonnet 5.5 mantle table we read on 4 October 2026 lists only us-gov-west-1 there. Plan new workloads on the EU geo profile, not on a single region.

IAM, logging and the eu. prefix are in our Claude on AWS Bedrock EU guide. Setting region="eu" on Google Cloud is in the Claude on Vertex AI in Europe guide.

How do you keep Claude Sonnet 5 inference in the EU?

Two documented routes keep Claude Sonnet 5 in EU member states: the Bedrock EU geo profile and Google Cloud’s Europe multi-region. The Claude API stores commercial data in the US and accepts only global and us. Foundry offers Global Standard and a US Data Zone, not a Europe zone. Foundry is the wrong pick if EU inference is mandatory.

AWS states on the Sonnet 5.5 card that the EU geo (eu.anthropic.claude-sonnet-5-5) keeps data within EU regions. The US geo keeps data in US and Canada regions. Global routes worldwide with no residency constraint. The same wording sits on the Sonnet 5 card for eu.anthropic.claude-sonnet-5. The cards list eu-central-1 (Frankfurt), eu-central-2 (Zurich), eu-north-1 (Stockholm), eu-south-1 (Milan), eu-south-2 (Spain), eu-west-1 (Ireland), eu-west-2 (London) and eu-west-3 (Paris). London and Zurich sit outside the EU. Call the profile from a member-state region, typically eu-central-1.

Google’s Sonnet 5.5 and Sonnet 5 pages list model availability and ML processing in the Europe multi-region, plus US multi-region and the global endpoint. Multi-region quota is 1,250 queries per minute, 12.5 million input tokens per minute and 1.25 million output tokens per minute. The global endpoint doubles those figures (2,500 QPM, 25 million input TPM, 2.5 million output TPM). Context length is 1,000,000 tokens. Set region="eu". The global endpoint doubles quota and leaves the EU.

Google’s retirement date for legacy Sonnet 5 is earlier than Anthropic’s: not sooner than 24 December 2026. Anthropic itself holds Sonnet 5 until at least 30 June 2027. If you still run Sonnet 5 on Google Cloud, plan the move to 5.5 before Google’s date. Sonnet 5.5 shares the Anthropic horizon on Google: not sooner than 28 September 2027.

Foundry remains the outlier. Microsoft’s hosting comparison places Sonnet 5 and Sonnet 5.5 on Hosted on Azure (Global Standard and US Data Zone) and on Hosted on Anthropic (Global Standard only). Anthropic is the processor on both paths. The four-provider matrix is in our Claude GDPR comparison.

How do you call Claude Sonnet 5.5 in the EU?

You need a cloud account in an EU region, the model ID with an EU prefix or region="eu", and you must not copy the global AWS sample. On Bedrock the Frankfurt call is eu.anthropic.claude-sonnet-5-5. On Google Cloud the ID stays claude-sonnet-5-5 and the region is eu. Five steps cover the first production call.

1. Create the AWS account or Google Cloud project in a member-state region. For Bedrock, eu-central-1 (Frankfurt) is the usual source region.
2. Enable the model. On Bedrock that is Model Access or AWS Marketplace (Sonnet 5.5 carries marketplace ID prod-pjfguoisodbd6). On Google Cloud that is Model Garden, model claude-sonnet-5-5.
3. Pin the EU ID in code. No generic sonnet, no global. prefix, no us-east-1.
4. Send a test prompt and read the response metadata. On Bedrock the modelId must start with eu.. On Google Cloud the region must be eu.
5. Set a cap: a budget alarm, a LiteLLM spend limit, or both. Token prices alone do not stop a runaway pipeline.

import boto3

client = boto3.client("bedrock-runtime", region_name="eu-central-1")
response = client.converse(
    modelId="eu.anthropic.claude-sonnet-5-5",
    messages=[{"role": "user", "content": [{"text": "Reply in one sentence from Frankfurt."}]}],
)
print(response["output"]["message"]["content"][0]["text"])

from anthropic import AnthropicVertex

client = AnthropicVertex(region="eu", project_id="YOUR_PROJECT")
message = client.messages.create(
    model="claude-sonnet-5-5",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Reply in one sentence from the Google EU region."}],
)
print(message.content[0].text)

Cache checkpoints on Sonnet 5.5 need at least 512 tokens. The legacy Bedrock card for Sonnet 5 sets the minimum at 1,024 tokens. Batch output up to 300,000 tokens needs the output-300k-2026-03-24 beta header. Leave temperature unset. When you migrate from Sonnet 5 to 5.5, replace the ID and change thinking.type from disabled to between_tools if you were turning thinking off.

FAQ

What does Claude Sonnet 5 cost in the EU?

Two US dollars input and ten US dollars output per million tokens on the Claude API, plus 10 percent on Bedrock EU geo and the Google Europe multi-region, so 2.20 / 11 USD. Batch is 1 / 5 USD globally and 1.10 / 5.50 USD on the 10 percent row. Cache reads are 0.20 USD globally and 0.22 USD on the 10 percent row. Foundry bills the same token rates in CCU at 0.01 USD. There is no published euro list price.

Claude Sonnet 5 or Claude Sonnet 5.5?

New workloads: Claude Sonnet 5.5. Same list price, newer knowledge cutoff (June 2026 versus January 2026), a longer Anthropic commitment (not sooner than 28 September 2027), and on Google Cloud a later date than Google’s Sonnet 5 horizon (24 December 2026). Stay on Sonnet 5 only while a wired eval or a vendor pin still requires the old ID. Plan the cutover before Google retires the legacy model.

Which Bedrock ID do you use from Frankfurt?

eu.anthropic.claude-sonnet-5-5 on bedrock-runtime in eu-central-1. The bare ID anthropic.claude-sonnet-5-5 is the card ID, not the on-demand modelId on bedrock-runtime. In-region on that endpoint is N/A. The global sample from us-east-1 with global.anthropic.claude-sonnet-5-5 leaves the EU. Legacy traffic still uses eu.anthropic.claude-sonnet-5.

Does Claude Sonnet 5 on Microsoft Foundry stay in the EU?

No. Foundry hosts Sonnet 5 and Sonnet 5.5 as Hosted on Azure (Global Standard and US Data Zone) and as Hosted on Anthropic (Global Standard only). There is no Europe data zone for Claude. Anthropic remains the processor. If EU inference is mandatory, use Bedrock or Google Cloud.

Is Claude Haiku 4.5 enough instead of Claude Sonnet 5?

For short classification and extraction, often yes. Haiku 4.5 costs 1 / 5 USD, has a 200,000 token context and a February 2025 knowledge cutoff. Anthropic holds the model no sooner than 15 October 2026. For a new pipeline that needs 2026 company knowledge, we stay on Sonnet 5.5. Haiku works as a cheap pre-filter in front of a Sonnet pass, not as the only default.

Is the Claude API still an option for EU companies?

As a chat and admin product, yes. As an inference location, no. inference_geo accepts only global and us. Workspace geo is us only today. Commercial data sits in the United States. For processing in EU member states, the Claude API is the wrong endpoint. Use it for Claude.ai seats, not for a production API that carries personal data.

Sources

1. Anthropic docs: Models overview (4 October 2026)
2. Anthropic docs: Claude Sonnet 5.5 (4 October 2026)
3. Anthropic docs: Claude Sonnet 5 (4 October 2026)
4. Anthropic docs: Pricing (4 October 2026)
5. Anthropic docs: Data residency (4 October 2026)
6. AWS Bedrock: Claude Sonnet 5.5 model card (4 October 2026)
7. AWS Bedrock: Claude Sonnet 5 model card (4 October 2026)
8. Google Cloud: Claude Sonnet 5.5 (4 October 2026)
9. Google Cloud: Claude Sonnet 5 (4 October 2026)
10. Google Cloud: Agent Platform pricing (4 October 2026)
11. Microsoft Learn: Claude hosting comparison (4 October 2026)