Anthropic
Operational· checked every few minutes

Claude Haiku 5.5 API

The Claude Haiku 5.5 API: Anthropic's fastest, lowest-priced model, 60% below list price.

Price per 1M tokens60%off
Input
$0.04$0.10
Output
$0.20$0.50
Input, long prompts
$0.20$0.50
Output, long prompts
$1.00$2.50
Cached input
$0.004
Cache writes (5 min)
$0.05$0.125
Cache writes (1 h)
$0.08$0.20
Get your API keyTry it in the playground
Model IDclaude-haiku-5-5
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-5-5",
    messages=[{"role": "user", "content": "Explain quicksort in two sentences."}],
    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
1M tokens
Max output
128K tokens
Reasoning
low to max
Input
Text, Image, PDF
Released
Oct 7, 2026
Status
Operational

Pricing

Claude Haiku 5.5 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
$0.04$0.10
Output
per 1M tokens
$0.20$0.50
Input, long prompts
per 1M tokens
$0.20$0.50
Output, long prompts
per 1M tokens
$1.00$2.50
Cached input
per 1M tokens
$0.004—
Cache writes (5 min)
per 1M tokens
$0.05$0.125
Cache writes (1 h)
per 1M tokens
$0.08$0.20

Prompts over 100K tokens cost 5 times every rate, cached input and cache writes included, the way Anthropic bills them. 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
$4.00
Anthropic list price
$10.00
You save every month$6.00

$72.00 a year

Overview

What is Claude Haiku 5.5?

Claude Haiku 5.5 is Anthropic's current Haiku, which Anthropic calls the cheapest, fastest and most capable small model it has released. The Zurelay Claude Haiku 5.5 API serves it through OpenAI-compatible and Anthropic-compatible endpoints at 60% below Anthropic's list price, for classification, extraction, routing, sub-agents and other high-volume work.

Claude Haiku 5.5 was released on October 7, 2026 and replaces Claude Haiku 4.5 as Anthropic's Haiku. Anthropic rates it the fastest model in its lineup and built it for high-volume, latency-sensitive work such as classification, extraction, routing and sub-agent tasks. Its list price is a tenth of Haiku 4.5's for prompts up to 100K tokens, and Anthropic says it costs about 75% less to run on average. On Zurelay, the Claude Haiku 5.5 API uses the model ID claude-haiku-5-5 and costs $0.04 per 1M input tokens and $0.20 per 1M output tokens.

The specs: a 1M-token context window, up from 200K on Haiku 4.5, up to 128K output tokens per request, text, image and PDF input, and text output. It is the first Haiku with adaptive thinking and the effort setting: thinking is on by default, and effort runs from low to max with medium as the default. Anthropic lists June 2026 as its reliable knowledge cutoff. It uses the tokenizer of Claude Opus 4.7 and later, so the same text counts as about 30% more tokens than on Haiku 4.5.

Anthropic's launch figures put it far ahead of Haiku 4.5: 1620 against 735 on the GDPval-AA v2.1 knowledge-work benchmark, where Sonnet 5.5 scores 1840; 72.4% against 15.7% on an offline subset of OSWorld 2.1 for computer use; and 39.2% against 0% on Terminal-Bench 4.0. Anthropic still recommends Sonnet 5.5 and Opus 5.5 for complex agentic coding, and Haiku 5.5 for narrower jobs such as summaries, compaction and sub-agent work.

Prompts over 100K tokens cost more, as on Anthropic's own price list: every rate is five times the base rate, so $0.20 input and $1.00 output per 1M tokens on Zurelay, against Anthropic's $0.50 and $2.50. A few API rules changed from Haiku 4.5: leave temperature, top_p and top_k unset, end the messages with a user turn instead of prefilling the reply, and set effort instead of a thinking budget. Prompt caching now works from a 512-token prompt, down from 4,096.

Strengths

Where Claude Haiku 5.5 shines. And what teams build with it.

01

Fastest in the lineup

Anthropic rates Haiku 5.5 its fastest model at standard speed, for replies a user is waiting on and agents that make many calls.

02

A tenth of Haiku 4.5's price

Its list price is a tenth of Haiku 4.5's for prompts up to 100K tokens, and Zurelay takes a further 60% off.

03

Close to Sonnet on knowledge work

Anthropic reports 1620 on GDPval-AA v2.1, against 735 for Haiku 4.5 and 1840 for Sonnet 5.5.

04

Built for sub-agents

Anthropic pitches it as the fast worker beside Opus 5.5 or Sonnet 5.5: the larger model plans, and Haiku 5.5 runs the many small steps.

Use cases

  • Classification and routing

    Tag, triage and route tickets, messages and documents at volume, where the cost of each call decides the budget.

  • Extraction

    Pull structured fields out of text, images and PDFs, with structured outputs or a tool schema.

  • Sub-agent workers

    Search, read, summarize and edit in parallel under a larger Claude model that plans the task and checks the results.

  • Live chat and support

    Customer-facing chat where a person waits on each reply. Start at low or medium effort and raise it only where answers need it.

Get started

Call Claude Haiku 5.5 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-5-5

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

    claude-haiku-5-5

Works with the tools you already use

Compare

Similar models. All on the same key.

55%off

Claude Sonnet 5.5

$0.90 / $4.50 per 1M tokens

Pick Claude Sonnet 5.5 when the work needs more reasoning, such as complex coding or long agent runs.

Claude Sonnet 5.5 API
46%off

Claude Haiku 4.5

$0.54 / $2.70 per 1M tokens

Pick Claude Haiku 4.5 if your pipeline relies on temperature, top_p or a thinking budget, which Haiku 5.5 no longer takes.

Claude Haiku 4.5 API
75%off

GPT-6 Luna

$0.025 / $0.125 per 1M tokens

Try GPT-6 Luna on the same Zurelay key for bulk text jobs, and compare the two on your own evals.

GPT-6 Luna API

FAQ

Claude Haiku 5.5 API questions.

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

How much does the Claude Haiku 5.5 API cost on Zurelay?

Zurelay charges $0.04 per 1M input tokens and $0.20 per 1M output tokens for prompts up to 100K tokens, with cached input at $0.004. Anthropic's list price is $0.10 input and $0.50 output, so you save 60%. Prompts over 100K tokens cost five times as much on every rate, as on Anthropic's own price list: $0.20 input and $1.00 output per 1M tokens.

Is Claude Haiku 5.5 cheaper than Claude Haiku 4.5?

Yes, by a wide margin. Anthropic's list price for Haiku 5.5 is a tenth of Haiku 4.5's for prompts up to 100K tokens and half of it above that, and Anthropic says it costs about 75% less to run on average. Its tokenizer counts about 30% more tokens for the same text, which the lower price more than makes up for.

How do I call Claude Haiku 5.5 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-5-5. Chat completions, streaming and tool calls use the request shapes you know from OpenAI, and reasoning_effort sets the effort level. Leave temperature, top_p and top_k unset: Haiku 5.5 rejects non-default sampling values.

Can I use Claude Haiku 5.5 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-5-5 as the model, or as the model Claude Code's Haiku alias points to.

What is the Claude Haiku 5.5 context window?

Claude Haiku 5.5 has a 1M-token context window, up from 200K on Haiku 4.5, and returns up to 128K output tokens per request. Prompts over 100K tokens are billed at five times the base rates, as Anthropic bills them, so keep routine prompts under that where you can.

Does Claude Haiku 5.5 support thinking?

Yes. It uses adaptive thinking, which is on by default, and the effort setting, from low to max with medium as the default: send output_config.effort in the Messages format or reasoning_effort on chat completions. You can turn thinking off with thinking type disabled at high effort or below. Anthropic's API rejects the older budget_tokens form on this model; Zurelay maps it to the nearest effort level instead. Thinking text comes back empty unless you set thinking display to summarized.

What is Claude Haiku 5.5 best at?

High-volume, latency-sensitive work: classification, extraction, routing, summaries, sub-agent tasks and live chat. Anthropic reports it close to Sonnet 5.5 on knowledge work, but still recommends Sonnet 5.5 or Opus 5.5 for complex agentic coding.

Does the Zurelay Claude Haiku 5.5 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 5.5 on Zurelay the same model as Anthropic's?

Yes. Your requests run on Claude Haiku 5.5 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