Anthropic
Operational· checked every few minutes

Claude Haiku 4.5 API

The Claude Haiku 4.5 API: fast Claude with extended thinking, 46% below list price.

Price per 1M tokens46%off
Input
$0.54$1.00
Output
$2.70$5.00
Cached input
$0.054$0.10
Cache-write pricesAll prices
Get your API keyTry it in the playground
Model IDclaude-haiku-4-5Updated
chat.py
from openai import OpenAI
client = OpenAI(
    base_url="https://api.zurelay.com/v1",
    api_key="YOUR_ZURELAY_KEY",
)
stream = client.chat.completions.create(
    model="claude-haiku-4-5",
    messages=[
        {"role": "user", "content": "Explain quicksort briefly."},
    ],
    stream=True,
)
for chunk in stream:
    if chunk.choices:
        print(chunk.choices[0].delta.content or "", end="")

Base URL https://api.zurelay.com/v1. Works with any OpenAI SDK, and the Anthropic SDK and Claude Code. Chat API docs

Context window
200K tokens
Max output
64K tokens
Reasoning
Thinking budget
Input
Text, Image
Released
Oct 15, 2025
Status
Operational

Pricing

Claude Haiku 4.5 API pricing. A fraction of the list price.

Pay as you go from prepaid credit, with no subscription. Top up from $10. Requests that fail are never billed.

RateZurelayAnthropic list
Input
per 1M tokens
$0.54$1.00
Output
per 1M tokens
$2.70$5.00
Cached input
per 1M tokens
$0.054$0.10
Cache writes (5 min)
per 1M tokens
$0.675$1.25
Cache writes (1 h)
per 1M tokens
$1.08$2.00

One rate at every prompt length, up to the full 200K context window. Streaming, tool calls and structured outputs cost nothing extra.

Every Claude model and price
Savings calculator

Move the sliders to your monthly usage.

Input tokens per month50M
Output tokens per month10M
Zurelay
$54.00
Anthropic list price
$100.00
You save every month$46.00

$552.00 a year

Cost

Claude Haiku 4.5 API cost, worked out. And compared.

Typical workloads at Zurelay's live prices next to the list price, and what other places charge for the same model.

  • Chatbot

    50,000 requests a month, 1,500 tokens in and 300 out, 30% of the input cached. $0.0014 a request.

    $70.07a month$129.75 at list

  • Documents

    20,000 requests a month, 8,000 tokens in and 500 out, 20% of the input cached. $0.00489 a request.

    $97.85a month$181.20 at list

  • Coding agent

    5,000 requests a month, 40,000 tokens in and 1,500 out, 85% of the input cached. $0.00913 a request.

    $45.63a month$84.50 at list

  • Classification

    1,000,000 requests a month, 400 tokens in and 5 out. $0.000229 a request.

    $229.50a month$425.00 at list

Where to buy Claude Haiku 4.5

Prices per 1M tokens

SellerInputOutput
Anthropic
List price, direct
$1.00$5.00
OpenRouter
List price, plus 5.5% on card top-ups
$1.00$5.00
Zurelay
No fee to add credit
$0.54$2.70

OpenRouter's fee as its pricing page listed it on October 28, 2026. Zurelay's prices are live. Some labs also discount batch or off-peak requests.

Overview

What is Claude Haiku 4.5?

Claude Haiku 4.5 is the Haiku Anthropic released in October 2025, which it described as its fastest model with near-frontier intelligence. The Zurelay Claude Haiku 4.5 API serves it through OpenAI-compatible and Anthropic-compatible endpoints at 46% below Anthropic's list price. It still fits pipelines built on it, especially ones that set temperature or a thinking budget; Claude Haiku 5.5 is the newer, lower-priced Haiku.

Claude Haiku 4.5 was released on October 15, 2025. Claude Haiku 5.5 replaced it as Anthropic's current Haiku on October 7, 2026, and Anthropic now lists Haiku 4.5 as a legacy model that is still available. At its launch, Anthropic described it as its fastest model with near-frontier intelligence that matched Claude Sonnet 4 on coding at one-third the cost and more than twice the speed. On Zurelay, the Claude Haiku 4.5 API uses the model ID claude-haiku-4-5 and costs $0.54 per 1M input tokens and $2.70 per 1M output tokens.

The specs: a 200K-token context window, up to 64K output tokens per request, text and image input, and text output. Its reliable knowledge cutoff is February 2025, with training data through July 2025. Haiku 4.5 supports extended thinking: set thinking type enabled with a budget_tokens value of at least 1,024 and below max_tokens. It has no effort setting and no interleaved thinking between tool calls. It tracks its remaining context window on its own, which Anthropic calls context awareness.

Anthropic reported 73.3% on SWE-bench Verified at launch and said Haiku 4.5 beats Claude Sonnet 4 at some tasks, such as using computers. It targets real-time, low-latency work like chat assistants, customer service agents and pair programming. Anthropic also pitches it as a sub-agent: a larger Claude model breaks a problem into steps, then a team of Haiku 4.5 instances completes the subtasks in parallel.

Haiku 4.5 is active, with retirement no sooner than October 15, 2026, and Anthropic has not announced a deprecation as of October 11, 2026. Haiku 5.5 has a tenth of its list price for prompts up to 100K tokens, but Haiku 4.5 still accepts temperature, top_p and top_k and a fixed thinking budget, which Haiku 5.5 does not. Prompt caching needs at least 4,096 tokens in the cached prompt, more than current Sonnet and Fable models require, so cache long, stable system prompts and documents.

Strengths

Where Claude Haiku 4.5 shines. And what teams build with it.

01

Fast replies

Built for speed at launch, a good match for replies a user is waiting on.

02

Sampling parameters still work

It accepts temperature, top_p and top_k, which Claude Haiku 5.5 and other newer Claude models reject, so tuned pipelines keep working.

03

Strong coding for its size

Anthropic reported 73.3% on SWE-bench Verified at launch, with coding on par with Claude Sonnet 4.

04

Built for sub-agents

Fast and cheap enough to run many copies in parallel under a larger Claude model that plans and reviews.

Use cases

  • Chat and support assistants

    Real-time chat and customer service agents where a user waits on every reply.

  • Sub-agent workers

    Parallel workers that search, read, extract or edit while a larger Claude model plans the task and checks the results.

  • Pair programming

    Quick completions, explanations and small fixes inside an editor, where latency matters more than depth.

  • High-volume extraction

    Tagging, routing, summarizing and pulling structured fields from large batches of text or images.

Get started

Call Claude Haiku 4.5 in three steps.

No waitlist and no new SDK. If your code already talks to OpenAI, it already talks to Zurelay.

  1. 1

    Create a key

    Sign up, add credit and create an API key. Set a monthly budget or a rate limit per key if you like.

    Get your API key
  2. 2

    Point your SDK at Zurelay

    Change the base URL. Everything else in your code stays the same.

    https://api.zurelay.com/v1
  3. 3

    Use claude-haiku-4-5

    Send chat completions as usual. Streaming, tool calls and usage reporting work as you expect.

    claude-haiku-4-5

Works with the tools you already use

Compare

Similar models. All on the same key.

60%off

Claude Haiku 5.5

$0.04 / $0.20 per 1M tokens

Pick Claude Haiku 5.5, Anthropic's current Haiku, for new work: a tenth of the list price for prompts up to 100K tokens, a 1M-token context window and effort control.

Claude Haiku 5.5 API
61%off

Claude Sonnet 5.5

$0.79 / $3.95 per 1M tokens

Pick Claude Sonnet 5.5 when you need more reasoning, adaptive thinking or a 1M-token context window, at a higher list price.

Claude Sonnet 5.5 API
65%off

Claude Opus 5.5

$1.40 / $7.00 per 1M tokens

Pick Claude Opus 5.5 to plan and review complex work that Haiku 4.5 sub-agents then carry out.

Claude Opus 5.5 API

FAQ

Claude Haiku 4.5 API questions.

Anything else? Ask our team and we’ll answer by email.

How much does the Claude Haiku 4.5 API cost on Zurelay?

Zurelay charges $0.54 per 1M input tokens and $2.70 per 1M output tokens, with cached input at $0.054. Anthropic's list price is $1.00 input and $5.00 output, so you save 46%. The rate is the same at every prompt length.

Is there a cheaper Claude API than Anthropic's?

Yes. Zurelay serves claude-haiku-4-5 at 46% below Anthropic's list price. It is the same model, paid from prepaid credit with no subscription. You keep your code and change the base URL and key. For new work, Claude Haiku 5.5 is also on Zurelay, at a much lower list price.

How do I call Claude Haiku 4.5 with the OpenAI SDK?

Create a Zurelay API key, set the base URL to https://api.zurelay.com/v1 and set the model to claude-haiku-4-5. Chat completions, streaming and tool calls use the request shapes you know from OpenAI, and temperature and top_p work as usual.

Can I use Claude Haiku 4.5 with the Anthropic SDK or Claude Code?

Yes. Set the Anthropic SDK's base URL, or ANTHROPIC_BASE_URL for Claude Code, to https://api.zurelay.com with no /v1, because the client adds /v1/messages itself. Authenticate with your Zurelay key and choose claude-haiku-4-5 as the model.

What is the Claude Haiku 4.5 context window?

Claude Haiku 4.5 has a 200K-token context window, roughly 150k words, and returns up to 64K output tokens per request. For larger inputs, use a model with a 1M-token window, such as Claude Sonnet 5.5.

Does Claude Haiku 4.5 support thinking?

Yes, in extended mode. In the Messages format, send thinking type enabled with a budget_tokens value of at least 1,024 and below max_tokens. Haiku 4.5 does not support adaptive thinking, the effort setting, or interleaved thinking between tool calls.

What is Claude Haiku 4.5 best at?

Fast, frequent calls: chat assistants, customer support, pair programming, sub-agents and bulk extraction, especially in pipelines that rely on temperature or a thinking budget. Anthropic reported coding on par with Claude Sonnet 4 at launch. For new high-volume work, Claude Haiku 5.5 is stronger and cheaper; for hard multi-step reasoning, a Sonnet or Opus model is a better fit.

Does the Zurelay Claude Haiku 4.5 API support streaming, and what are the limits?

Yes. Streaming works on chat completions and on the Messages API; on chat completions, set stream_options.include_usage to get token counts at the end of the stream. Each key can have its own monthly budget and requests-per-minute cap. If a route fails, Zurelay retries on another route, and failed requests are never billed.

Is Claude Haiku 4.5 on Zurelay the same model as Anthropic's?

Yes. Your requests run on Claude Haiku 4.5 itself. Zurelay doesn't swap in a different or smaller model. Only smart routing, which you turn on yourself, can answer with a close alternative when the model is down, and the x-zurelay-model header names the model that answered. Output is sampled, so exact wording varies between calls on any API. Zurelay is an independent service and is not affiliated with or endorsed by Anthropic.

More models on Zurelay

All models

Stop paying list price.
Start saving today.

Create a key in seconds, point the SDK you already use at Zurelay, and every request costs up to 90% less from the first token.

Get your API keyTalk to an engineer