Claude Haiku API Example

How to use claude-haiku-4-5 via the Anthropic Python SDK. Working code for classification, tagging, extraction, and high-volume tasks where Haiku's low cost matters.

claude-haiku-4-5-20251001 is the go-to model for high-volume, cost-sensitive workloads. Same API, 3.75× cheaper than Sonnet.

Basic Haiku call

import anthropic

client = anthropic.Anthropic()

response = client.messages.create(
    model="claude-haiku-4-5-20251001",  # ← the key change
    max_tokens=256,
    messages=[
        {"role": "user", "content": "Classify this support ticket: 'My account was charged twice.' Category: billing, technical, account, other"}
    ]
)
print(response.content[0].text)  # "billing"

Model comparison table

Batch classification with Haiku (cost-optimised)

import anthropic
from concurrent.futures import ThreadPoolExecutor

client = anthropic.Anthropic()

CATEGORIES = ["billing", "technical", "account", "shipping", "other"]

def classify_ticket(ticket: str) -> str:
    response = client.messages.create(
        model="claude-haiku-4-5-20251001",
        max_tokens=32,
        system=f"Classify the support ticket into exactly one of: {', '.join(CATEGORIES)}. Reply with only the category name.",
        messages=[{"role": "user", "content": ticket}]
    )
    return response.content[0].text.strip().lower()

tickets = [
    "I was charged twice on my credit card",
    "The app keeps crashing on iOS 17",
    "I forgot my password and can't reset it",
    "My package hasn't arrived in 3 weeks",
]

with ThreadPoolExecutor(max_workers=5) as pool:
    results = list(pool.map(classify_ticket, tickets))

for ticket, category in zip(tickets, results):
    print(f"{category:12} | {ticket}")

Structured extraction with Haiku + tool use

import anthropic, json

client = anthropic.Anthropic()

extract_tool = {
    "name": "extract_entities",
    "description": "Extract named entities from text",
    "input_schema": {
        "type": "object",
        "properties": {
            "people": {"type": "array", "items": {"type": "string"}},
            "organizations": {"type": "array", "items": {"type": "string"}},
            "dates": {"type": "array", "items": {"type": "string"}}
        },
        "required": ["people", "organizations", "dates"]
    }
}

response = client.messages.create(
    model="claude-haiku-4-5-20251001",
    max_tokens=512,
    tools=[extract_tool],
    tool_choice={"type": "tool", "name": "extract_entities"},
    messages=[{
        "role": "user",
        "content": "Elon Musk founded SpaceX in 2002 and acquired Twitter in October 2022."
    }]
)

for block in response.content:
    if block.type == "tool_use":
        print(json.dumps(block.input, indent=2))
# Output:
# {
#   "people": ["Elon Musk"],
#   "organizations": ["SpaceX", "Twitter"],
#   "dates": ["2002", "October 2022"]
# }

When Haiku is fast enough

• Single-label classification (sentiment, intent, category): Haiku matches Sonnet quality on simple <5-class tasks.
• Named entity extraction: Structured extraction via tool use is near-identical quality.
• Short summarization (<100 word output): Quality gap vs Sonnet is minimal.
• Keyword/tag extraction: Haiku performs at parity with Sonnet.
• Translation (common language pairs): Quality is strong for major languages.

Use the Cost Calculator to model Haiku vs Sonnet costs at your volume. See Python quickstart for the full API setup guide.

Frequently asked questions

Free tools

More examples