Anthropic Classification and extraction
Claude Haiku 5.5
Handle repeatable requests with a clear answer shape and a review path for exceptions.
Anthropic positions Claude Haiku 5.5 for tasks such as classification, routing and extraction. Use it on TokenHot to draft a queue choice, extract named fields or summarize updates for review.
- Text and image input
- Text output
- 1M-token catalog context
TokenHot's public API example uses text in Messages. The chat Playground accepts one new image attachment per message. Anthropic's adaptive thinking settings describe its own API, so confirm parameter support for your integration.
AI Chat
Send a prompt and see the response here.
claude-haiku-5-5Start a conversation with this model and ask follow-up questions.
USD /1M tokens
Pricing
| Group | Input | Output | Cache Read | Cache Write |
|---|---|---|---|---|
| default | $0.5 | $2.5 | $0.012 | $0.15 |
| claude-0.34 | $0.17 | $0.85 | $0.00408 | $0.051 |
Public list prices are shown in USD. Final charges may vary by account group and usage tier.
Overview
What is Claude Haiku 5.5?
Start with a narrow task, a known set of acceptable values and instructions for cases that do not fit. Keep unknown items visible so the model has an alternative to guessing.
Track acceptance and retries on a representative sample before sending a full queue. Measure cost per accepted result on the same inputs when comparing routes.
Route tickets by an explicit queue policy
Classify one ticket using a small, named set of queues. Ask for a reason tied to the ticket and an escalation flag for ambiguous cases.
Extract named fields from a form excerpt
Provide the field names, an example and the form excerpt. Ask the model to leave a field empty when the source does not establish its value.
Summarize a batch of short issue updates
Supply a set of brief issue updates with IDs. Request a compact digest that keeps unresolved items visible and avoids assigning owners the source never names.
From brief to deliverable
Put Claude Haiku 5.5 to work
Support operations
Route tickets by an explicit queue policy
Classify one ticket using a small, named set of queues. Ask for a reason tied to the ticket and an escalation flag for ambiguous cases.
- Tools you need
- Labeled inputs, an evaluation checklist and a reviewer.
- Keep in mind
- Keep the final routing decision in your application.
View a task brief
Read the ticket and routing policy. Return one allowed queue, one supporting clue and an escalation flag. If the policy leaves the case ambiguous, mark it for human review.
Data operations
Extract named fields from a form excerpt
Provide the field names, an example and the form excerpt. Ask the model to leave a field empty when the source does not establish its value.
- Tools you need
- Labeled inputs, an evaluation checklist and a reviewer.
- Keep in mind
- Use a validator before writing any record.
View a task brief
Extract only the requested fields. For each value, identify the matching label or source line. Leave it empty when the form does not establish a value.
Team coordinators
Summarize a batch of short issue updates
Supply a set of brief issue updates with IDs. Request a compact digest that keeps unresolved items visible and avoids assigning owners the source never names.
- Tools you need
- Labeled inputs, an evaluation checklist and a reviewer.
- Keep in mind
- Keep the original issue IDs for review.
View a task brief
Group the supplied updates into completed, active and unresolved work. Preserve issue IDs. List unknown owners and dates as open rather than inferring them.
Capabilities in context
Key features
| Feature | What you get | Why it matters |
|---|---|---|
| TokenHot model ID | claude-haiku-5-5 | Use this exact value in the model field. |
| Catalog modalities | Text and image input, text output | Use a supplied image as evidence and ask for a written result. |
| Catalog context | 1M tokens | Send the relevant material and leave room for the answer. |
| TokenHot API example | Text request through /v1/messages | Start with the published request and confirm additional fields for your endpoint. |
| Anthropic adaptive thinking | On by default on the Claude API | Anthropic documents effort settings for its API. Check what the TokenHot endpoint exposes before relying on those settings. |
TokenHot's catalog and request example describe this route. Anthropic's model documentation describes the Claude API. Scenario art was made with ImageGen to illustrate the workflows. TokenHot model details · Anthropic Claude Haiku 5.5 notes · Anthropic thinking configuration
Audience & fit
Where Claude Haiku 5.5 fits
Support operations
A queue recommendation and an escalation flag.
Data operations
A field checklist with source evidence and missing values.
Team coordinators
A status digest grouped by completed, active and unresolved work.
Cost context
Plan the cost of a useful result
Check current rates
Use the TokenHot pricing table for the model and account group.
Measure accepted work
Include retries and review when comparing the cost of a usable result.
Limit the evidence packet
Keep the source material relevant and specify a response format sized to the task.
Use the current pricing table for available rates and account groups. Model IDs and prices come from TokenHot’s catalog; provider pricing and subscriptions are separate.
API
Code examples
curl --location --request POST 'https://api.tokenhot.ai/v1/messages' \
--header 'Authorization: Bearer <token>' \
--header 'Content-Type: application/json' \
--data-raw '{
"model": "claude-haiku-5-5",
"messages": [
{
"role": "user",
"content": "What can you do?"
}
],
"max_tokens": 100
}'Blog
Related Articles
FAQ
Frequently asked questions
What is Claude Haiku 5.5 best suited for?
Claude Haiku 5.5 is suited to the input, output, and capability types listed on this page. Test production workloads before choosing it for a critical system.
How is Claude Haiku 5.5 priced?
Pricing is shown in the pricing section above. Actual cost depends on usage volume, account group, and the request payload.
How can I call Claude Haiku 5.5?
Use the model ID shown above with one of the supported API protocols. Code examples are provided when usage data is available.
Which model ID do I use for Claude Haiku 5.5?
Use claude-haiku-5-5 in the model field of the TokenHot request.
What does the published TokenHot API example show?
The page shows a text request to /v1/messages. Confirm additional parameters and image payloads for the endpoint before integrating them.
Can I attach a screenshot in the Playground?
Yes. The chat Playground accepts one new image attachment per message and returns a written response.
Can I configure adaptive thinking?
Anthropic documents adaptive thinking and effort on its Claude API. Check parameter availability for the TokenHot endpoint you use.
When should I compare with Claude Sonnet 5.5?
Compare the same ambiguous tickets or extraction cases on Sonnet if the task requires judgments across several sources.
Get started
Build with Claude Haiku 5.5
API