Claude Haiku 4.5 API pricing

By the aitokensai editors · Verified 2026-11-01 Sources: Anthropic official pricing · Anthropic docs

Ultra-fast classification, extraction and tight loops. Available at about 50% below official list.

USD, standard tier. List price cross-checked against Anthropic’s published pricing on 2026-11-01; the discounted route was read from the provider’s live rate config the same day.

Specifications

What Claude Haiku 4.5 costs at your volume

Calculated at $0.5 input / $2.5 output per 1M tokens, 5.0× apart on this model, assuming a 5:1 input-to-output ratio. That ratio is the single largest source of error in this table: resend-heavy workloads drift well past it, and at 5.0× the correction is not linear. The cost calculator takes the actual figures.

When Claude Haiku 4.5 is the right call

At $0.5 in and $2.5 out per million tokens, Claude Haiku 4.5 lands where most production deployments actually live — cheap enough to leave running, expensive enough that you still look at the bill. The 5.0× output multiple is conventional, so standard cost intuition holds: output dominates, context matters, neither is pathological.

It is not the answer at either end. Bulk classification finds identical results several times cheaper, and the hardest reasoning problems reward moving up rather than staying here. Everything in between is this model's ground.

Where Claude Haiku 4.5 sits in the index

Claude Haiku 4.5 is 106 of 226 on input rate and 136 of 226 on output rate. A gap of 30 places, because on this model the output rate is 5.0× the input rate, steeper than the index median of 4.0× An output-heavy workload therefore pushes it 30 places down the table, which is why the cheapest model per input token is routinely not the cheapest per task. Percentile is the share of the 226 models in this index that cost more than Claude Haiku 4.5.

On an agent mix the lowest blended rate in the same index is Gemma2 9B It at $0.01 per 1M tokens, about 200.0× less than this model at the same ratio. Across the whole index, sorting by input price alone moves a model by as much as 74 places; on Claude Haiku 4.5 specifically the shift is 30 — which is what the re-ranked index exists to show.

Claude Haiku 4.5 is 106 of 226 across every model in the index; the comparison that usually settles a choice is Anthropic's own lineup further down.

Claude Haiku 4.5 among Anthropic's 9 models

Claude Haiku 4.5 is the 2nd cheapest of Anthropic's 9 models in this index at $0.5 per 1M input tokens. The range inside that group runs from Claude Haiku 5.5 at $0.05 to Claude Fable 5 at $5 (a spread of 100×, and this one sits at 10).0× the group's cheapest tier.

On input rate alone, stepping up to Claude Sonnet 5 costs 2.0× the input rate and dropping to Claude Haiku 5.5 cuts it by 90%. Both measured at 2026-11-01, the same check as every other rate on this page.

Full Anthropic range, including models outside this shortlist: Anthropic rate index.

Similarly priced alternatives

Blended comparison (input + output per 1M). Price is one axis These are the three closest to Claude Haiku 4.5 in blended rate, not interchangeable substitutes for it.

Calling Claude Haiku 4.5

The identifier you send is claude-haiku-4-5-20251001. Copy it exactly, because a mismatch returns a model error rather than a connection error, which sends most people to debug the wrong thing. Everything else is a base URL argument (Claude Haiku 4).5's own SDKs work unchanged.

from openai import OpenAI

client = OpenAI(base_url="https://aicomp.ai/v1", api_key="sk-...")

resp = client.chat.completions.create(
    model="claude-haiku-4-5-20251001",
    messages=[
        {"role": "system", "content": "Answer in one short paragraph."},
        {"role": "user", "content": "Explain what changed in this release."},
    ],
    max_tokens=300,
)

# prompt_tokens / completion_tokens 是真实账单的唯一依据 —— 记下来，别猜
print(resp.usage.prompt_tokens, resp.usage.completion_tokens)

Log the usage object on every call, not just while you are optimising. On Claude Haiku 4.5 output costs 5.0× what input costs, so the same monthly token count bills very differently depending on how much of it comes back as output — measure your own split before extrapolating from the headline input rate, because rates move with vendor promotions.

Setting it up

Pointing an existing integration at Claude Haiku 4.5 is a base URL change, not a rewrite. Pick your tool:

• Claude Code — set the base URL via environment variables
• Cursor — override the API endpoint in settings
• OpenAI / Anthropic SDK — one base_url argument
• Python, sync, streaming and async examples
• Cutting the bill — eight levers ranked by effect

What different usage shapes cost on Claude Haiku 4.5

A monthly token budget does not convert cleanly into a monthly invoice, and Claude Haiku 4.5 is an ordinary case rather than an extreme one: output is priced at 5.0× input, above the median of 4.0× across the 226 models in our index. Everything follows from that one number At 5:1 it is manageable, at 1:1 the same token count costs half as much. Below are five shapes real Claude Haiku 4.5 deployments fall into, each costing it out at the split it actually produces.

The cheapest comparison is GPT-5 nano at $0.025 / $0.2 per 1M — 13.3× cheaper than Claude Haiku 4.5 on a blended rate. Treat it as a rate floor, not a recommendation: it says nothing about whether GPT-5 nano can do the work you need Claude Haiku 4.5 for.

Read the table as a sanity check rather than a forecast. If your own output share sits anywhere near 3%, everything above is priced almost entirely at Claude Haiku 4.5's $0.5 input rate; at 40% the $2.5 output rate has taken the larger half. Nobody's split is exactly the one you assumed when you picked the model, which is why the calculator asks for it before quoting anything.

FAQ

Get Claude Haiku 4.5 access