Model details
claude-haiku-4-5 API Pricing and Integration
View current input, output and cache prices for claude-haiku-4-5 on ZHUAN and see how to send a request with its model ID.
Current prices
claude-haiku-4-5Current prices for new keys · USD per million tokens
- Input
- $0.296546
- Output
- $1.479335
- Cache read
- $0.029672
- Cache write
- $0.369834
Daily sync · 23 models ·
How do I call claude-haiku-4-5?
Use claude-haiku-4-5 as the request model ID. The OpenAI-compatible endpoint is https://key.zhuan.xyz/v1/chat/completions, with a key from the ZHUAN console. The example reads ZHUAN_API_KEY from the environment; generation incurs usage charges.
import os
from openai import OpenAI
client = OpenAI(
api_key=os.environ["ZHUAN_API_KEY"],
base_url="https://key.zhuan.xyz/v1",
max_retries=0,
)
response = client.chat.completions.create(
model="claude-haiku-4-5",
messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)
Anthropic Messages example
Set the Anthropic Python SDK base_url to https://key.zhuan.xyz/proxy. This example limits output to 128 tokens. Adjust it for your task and validate supported parameters first.
import os
from anthropic import Anthropic
client = Anthropic(
api_key=os.environ["ZHUAN_API_KEY"],
base_url="https://key.zhuan.xyz/proxy",
max_retries=0,
)
message = client.messages.create(
model="claude-haiku-4-5",
max_tokens=128,
messages=[{"role": "user", "content": "Hello"}],
)
for block in message.content:
if block.type == "text":
print(block.text)
How should I evaluate this model for my tasks?
Try a set of real tasks and compare output quality, token usage and latency before choosing a model.
How do I check request charges?
This page shows prices for newly created keys, in USD per million tokens, from the same daily-synced data as the price catalog. Existing key prices are shown in the console. After a request, check the model, input, output and cache tokens and the actual charge in request history. The public catalog does not promise a fixed per-request cost.
What should I check before integration?
- The catalog and prices change. A catalog entry does not guarantee availability for every request; check the status and request ID if a model is unavailable.
- Context, output limits, concurrency and advanced parameters depend on actual service responses. Validate a small request before increasing load.
- Retries, timeouts and interrupted streams may still incur usage. Contact support with the request ID if a charge appears incorrect.