What is the batch price?

50 percent of standard API prices, input and output.

What is the typical vs maximum time?

Most batches finish within 1 hour; expire if not done in 24 hours.

What is the batch size limit?

100,000 requests or 256 MB, whichever comes first.

What is the customid format?

1-64 chars, ^[a-zA-Z0-9-]{1,64}$, unique within the batch.

What is the result order?

Not guaranteed; match by customid.

What is the result types?

succeeded, errored, canceled, expired.

What is the not billed?

errored, canceled and expired requests.

Prompt Engineering & Structured Output20% of the exam

Message Batches API and batch strategy

When the Message Batches API's 50% discount is worth its up-to-24-hour window, how to correlate results by custom_id, resubmit only failed requests, size submission intervals against an SLA, and why client tools cannot run inside a batched request.

Official exam objective

Design efficient batch processing strategies

~8 min read · 10 practice questions · updated Oct 22, 2026

What the exam expects

Match the API to the workflow's latency need: synchronous Messages API for blocking work such as pre-merge checks, the Message Batches API for overnight or weekly analysis.
Build batch requests with meaningful custom_id values and correlate results by custom_id, never by position.
Handle batch failures by resubmitting only the failed custom_ids, each with the change its failure needs (chunk oversized documents, resend expired requests unchanged).
Calculate how often to submit batches so a worst-case batch still meets an SLA.
Explain why a batched request cannot run your tools mid-request, and design around it.
Refine the prompt on a sample before batch-processing large volumes.

Key terms: Message Batches API (asynchronous processing of many Messages requests at 50% of standard prices); custom_id (your identifier for each request, echoed on its result); result type (succeeded, errored, canceled or expired); processing window (up to 24 hours before unfinished requests expire); SLA arithmetic (submission interval + batch window + downstream time).

Key facts

Fact	Detail	Source
Batch price	50 percent of standard API prices, input and output	Batch processing (opens in a new tab)
Typical vs maximum time	Most batches finish within 1 hour; expire if not done in 24 hours	Batch processing (opens in a new tab)
Batch size limit	100,000 requests or 256 MB, whichever comes first	Batch processing (opens in a new tab)
Oversized batch	413 `request_too_large`	Batch processing (opens in a new tab)
`custom_id` format	1-64 chars, `^[a-zA-Z0-9_-]{1,64}$`, unique within the batch	Create a Message Batch (opens in a new tab)
Result order	Not guaranteed; match by `custom_id`	Retrieve Message Batch results (opens in a new tab)
Result types	`succeeded`, `errored`, `canceled`, `expired`	Batch processing (opens in a new tab)
Not billed	`errored`, `canceled` and `expired` requests	Batch processing (opens in a new tab)
One failure	Does not affect other requests in the batch	Batch processing (opens in a new tab)
Submitted batch	Cannot be modified; cancel and submit a new one	Batch processing (opens in a new tab)
Client tool call	Request ends with `stop_reason` `tool_use`; your code must send the `tool_result`	Handle tool calls (opens in a new tab)

Batch price

Detail: 50 percent of standard API prices, input and output
Source: Batch processing (opens in a new tab)

Typical vs maximum time

Detail: Most batches finish within 1 hour; expire if not done in 24 hours
Source: Batch processing (opens in a new tab)

Batch size limit

Detail: 100,000 requests or 256 MB, whichever comes first
Source: Batch processing (opens in a new tab)

Oversized batch

Detail: 413 request_too_large
Source: Batch processing (opens in a new tab)

custom_id format

Detail: 1-64 chars, ^[a-zA-Z0-9_-]{1,64}$, unique within the batch
Source: Create a Message Batch (opens in a new tab)

Result order

Detail: Not guaranteed; match by custom_id
Source: Retrieve Message Batch results (opens in a new tab)

Result types

Detail: succeeded, errored, canceled, expired
Source: Batch processing (opens in a new tab)

Not billed

Detail: errored, canceled and expired requests
Source: Batch processing (opens in a new tab)

One failure

Detail: Does not affect other requests in the batch
Source: Batch processing (opens in a new tab)

Submitted batch

Detail: Cannot be modified; cancel and submit a new one
Source: Batch processing (opens in a new tab)

Client tool call

Detail: Request ends with stop_reason tool_use; your code must send the tool_result
Source: Handle tool calls (opens in a new tab)

Sources for this section (4)

Check yourself

Why is the Message Batches API wrong for a check that blocks merging?
When a batched request calls a client tool, what happens?
What are the per-batch size limits?
When is it safe to resubmit a failed request unchanged?

The questions are free to read. Answers with their sources and self-grading are included with the pack.

Keep reading with the Claude Certified Architect - Foundations pack

Core concepts
Exam traps
Worked scenario

The pack includes:

All 30 lessons of the Claude Certified Architect - Foundations guide, with reading progress
Interactive decision walks and step-by-step sequence steppers
Inline practice questions and one-click topic practice (10 questions for this lesson)
An official source cited for every claim
An AI tutor that answers from the guide
Unlimited timed practice exams

Get the pack Already own it? Sign in

Practice for the Claude Certified Architect - Foundations exam

Certifyforce is an independent study resource and is not affiliated with, endorsed by or sponsored by Anthropic. Claude is a trademark of Anthropic.