Claude Haiku 4.5 pricing
$1.00 per million input tokens, $5.00 per million output. 200K token context window. Read from Anthropic’s own documentation on August 3, 2026.
- Input
- $1.00 per 1M tokens
- Cached input
- $0.1000 10% of base
- Output
- $5.00 5.0× input
- Context window
- 200K verified
- Token counting
- Estimate o200k_base
- Price verified
- 2026-08-03 2 days ago
Source: Anthropic pricing documentation. Prices change without notice — verify before committing spend.
What Claude Haiku 4.5 costs on real work
Four workload shapes at 100,000 requests a month. The point of showing four is that the ranking between models changes depending on which one describes you.
| Workload | In | Out | Per request | Per month |
|---|---|---|---|---|
| ClassificationShort input, one-word answer. Input-dominated. | 500 | 50 | $0.000750 | $75.00 |
| Chat turnA system prompt plus a few turns of history. | 1,500 | 300 | $0.003000 | $300.00 |
| Document summaryA long document in, a paragraph out. | 20,000 | 800 | $0.0240 | $2,400.00 |
| Code generationOutput-heavy — where output pricing dominates. | 2,000 | 1,500 | $0.009500 | $950.00 |
Put your own numbers in the cost calculator, or measure a real prompt first in the token counter. If your requests share a stable prefix, the cached rate applies to most of your input — check the structure in the cache checker.
Counting tokens for Claude Haiku 4.5
Anthropic does not publish a tokenizer that runs in a browser, so any pre-flight count for Claude Haiku 4.5 is an estimate rather than a measurement.
Anthropic does not publish a client-side tokenizer. Counted with o200k_base and scaled ~1.18x, the commonly reported gap between tiktoken and Claude's pre-4.7 tokenizer.
Treat it as accurate to within roughly ten to twenty percent. That is fine for budgeting and wrong for sizing a prompt right at a context window boundary — where precision matters, use Anthropic’s own token counting endpoint from your backend. The methodology page sets out every scaling factor used here.
Other Anthropic models
The tier question: is a cheaper model in the same family enough for your task?
| Model | Input | Output | Context | Chat turn |
|---|---|---|---|---|
| Claude Haiku 4.5 — this page | $1.00 | $5.00 | 200K | $0.003000 |
| Claude Fable 5 | $10.00 | $50.00 | 1M | $0.0300 |
| Claude Opus 5 | $5.00 | $25.00 | 1M | $0.0150 |
| Claude Opus 4.8 | $5.00 | $25.00 | 1M | $0.0150 |
| Claude Opus 4.6 | $5.00 | $25.00 | 1M | $0.0150 |
| Claude Sonnet 5 | $2.00 | $10.00 | 1M | $0.006000 |
| Claude Sonnet 4.6 | $3.00 | $15.00 | 1M | $0.009000 |
Alternatives from other providers
Models priced nearest to Claude Haiku 4.5, not the cheapest on the market — those are the ones actually worth evaluating against it.
- OpenAIGPT-5.4 mini$0.7500 in · $4.50 out↓ 17% on a chat turn
- xAIGrok 4.3$1.25 in · $2.50 out↓ 13% on a chat turn
- xAIGrok Build 0.1$1.00 in · $2.00 out↓ 30% on a chat turn
- Mistral AIMagistral Medium$2.00 in · $5.00 out↑ 50% on a chat turn
Side-by-side comparisons: Claude Haiku 4.5 vs GPT-5 mini · Claude Haiku 4.5 vs Gemini 3.5 Flash
Frequently asked questions
- How much does Claude Haiku 4.5 cost?
- $1.00 per million input tokens and $5.00 per million output tokens, with cached input at $0.1000 per million. On a typical chat turn of 1,500 input and 300 output tokens that is $0.003000 per request, or $300.00 per month at 100,000 requests. Read from Anthropic's own documentation on August 3, 2026.
- Can I count Claude Haiku 4.5 tokens exactly?
- No. Anthropic does not publish a tokenizer that runs in a browser, so any pre-flight count for Claude Haiku 4.5 is an estimate. Anthropic does not publish a client-side tokenizer. Counted with o200k_base and scaled ~1.18x, the commonly reported gap between tiktoken and Claude's pre-4.7 tokenizer. Treat it as accurate to within roughly ten to twenty percent and never as the basis for sizing a prompt right at a context window boundary.
- What is the context window of Claude Haiku 4.5?
- 200,000 tokens. That budget covers everything in the request — system prompt, conversation history, tool definitions, documents — plus the response itself, not just your input.
- Why is output more expensive than input on Claude Haiku 4.5?
- Output costs 5.0 times input here. Input is processed in a single parallel pass, while output is generated one token at a time with a full pass over the model for each. That is why a model that answers concisely can be cheaper in production than one with a lower headline rate.