LEGACY HAIKU200K CONTEXTTEXT + VISIONPROMPT CACHINGBATCH -50%
Claude Haiku 4.5 API Pricing
Claude Haiku 4.5 was Anthropic's lowest-cost model from its release on October 15, 2025 until Claude Haiku 5.5 replaced it on October 7, 2026. Anthropic now labels it legacy and recommends migrating, with retirement not sooner than October 15, 2026. It still bills $1/M input, $5/M output and $0.10/M cache reads at any prompt length, with a 200K-token window and text and image input. Checked by hand against platform.claude.com.
Input · per 1M tokens
$1.00/M
Unchanged since launch ↔ flat
Output · per 1M tokens
$5.00/M
Unchanged since launch ↔ flat
Cached input · 90% off
$0.10/M
Cache 5min or 1h −90%
Effective · agentic blend
$0.682/M
92/8 split · 82% cache
§ 01 / TERMINAL
Run the numbers.
Live calculator pre-loaded with current Claude Haiku 4.5 rates. Tweak amount, workload split, or cache hit rate — share the URL to share the calculation.
$ /mo
Workload split
Prompt cache hit rate
Tokens you can process
—
Words equivalent (English)
—
Effective rate
—
§ 02 / SCENARIOS
Real-world presets.
CLASSIFICATION
Lead scoring classifier
$0.00096/record
CUSTOMER SUPPORT
Support bot first reply
$0.00348/ticket
BULK RAG
Knowledge base answering
$0.00718/answer
LIGHT AGENT
Back-office data cleanup
$0.016/task
Estimate · anthropic-bpe-estimate · ≈3.5 chars/token Auto-counts as you type
This is a chars-per-token approximation, not a real tokenizer. Actual tokens vary by language, code density, and tool-call overhead — counts are typically ±10–20% off for English prose, more for code or non-Latin scripts. For exact billing, use the vendor's official tokenizer.
Characters 562
Words 96
Tokens (estimated) 161 tokens
Cost as input · uncached $0.00016 USD
Cost as output · uncached $0.00081 USD
Cost as cached input $0.00002 USD
| Model | Input /M | Output /M | Effective blended | Context | Best for |
|---|---|---|---|---|---|
| Claude Haiku 4.5 Current | $1.00 cache $0.10 | $5.00 | $0.682 agentic 92/8 | 200K | Classification · support automation |
| Claude Sonnet 4.6 | $3.00 cache $0.30 | $15.00 | $2.05 +3.0× | 1M | Production agents · coding |
| Claude Sonnet 5.5 | $2.00 cache $0.10 | $10.00 | $1.29 +89% | 1M | Newest Sonnet · production agents |
| Claude Haiku 5.5 | $0.10 cache $0.01 | $0.50 | $0.0682 −10× | 1M | The successor, 90% cheaper on short prompts |
| GPT-5.4 nano | $0.20 cache $0.02 | $1.25 | $0.148 −4.6× | 400K | Low-cost routing · simple extraction |
| Gemini 3.5 Flash-Lite | $0.30 cache $0.03 | $2.50 | $0.272 −2.5× | 1M | High-volume agents · translation |
| DeepSeek V4.1 Flash | $0.30 cache $0.006 | $1.20 | $0.15 −4.5× | 1M | Ultra-cheap bulk generation |
| DeepSeek V4 Pro | $1.32 cache $0.044 | $3.96 | $0.569 −17% | 1M | Budget reasoning · tool use |
Frequently asked.
Practical pricing questions for Claude Haiku 4.5, now the legacy Haiku behind Haiku 5.5.
Q · 01 Is Claude Haiku 4.5 the cheapest Claude model? +
Not since October 7, 2026. Claude Haiku 5.5 lists at
$0.10/M input and $0.50/M output for prompts up to 100,000 tokens - a tenth of Haiku 4.5's $1/M and $5/M - and at half of them above 100,000 tokens. Haiku 4.5 has one card at any prompt length. Under a 92/8 input-output workload with 82% cache hits its blended cost is $0.68/M, against $0.0682/M for Haiku 5.5 on short prompts.Q · 02 How does prompt caching work for Haiku 4.5? +
Cache hits are billed at
$0.10/M, which is 10% of the $1/M base input price. Cache writes cost $1.25/M for a 5-minute cache and $2/M for a 1-hour cache. Use caching for stable system prompts, policy text, retrieved context, and repeated instructions.Q · 03 When was the last price change? +
We found no change to Haiku 4.5's own prices since launch. Anthropic released it on October 15, 2025, and the live pricing table still shows
$1/M input, $0.10/M cache hits and $5/M output as of October 9, 2026. What changed is its place in the lineup: with Haiku 5.5's launch on October 7, 2026, Anthropic labels Haiku 4.5 a legacy model and recommends migrating; its retirement date is not sooner than October 15, 2026.Q · 04 Is there a Batch API discount? +
Yes. Anthropic's Batch API applies a 50% discount to both input and output tokens. For Claude Haiku 4.5, the batch table lists
$0.50/M input and $2.50/M output, compared with the standard $1/M and $5/M rates.Q · 05 Does Haiku 4.5 have regional pricing differences? +
The first-party Claude API uses global routing at standard list price by default. Anthropic says regional and multi-region partner endpoints on Bedrock and Vertex AI include a 10% premium for Claude 4.5 models and later. The first-party
inference_geo 1.1x data-residency multiplier applies to Claude Opus 4.6, Claude Sonnet 4.6, and later models, not Haiku 4.5.Q · 06 Does Haiku 4.5 support vision? +
Yes. Anthropic's model page lists Claude Haiku 4.5's input and output as text and images in, text out, with a 200K-token context window and 64K max output. Images bill as input tokens at the same rates as text. Its successor, Claude Haiku 5.5, also takes images, with a 1M-token window.
Q · 07 Are volume discounts published? +
No public volume-discount ladder is published for Haiku. Anthropic says volume discounts may be available for high-volume users and are negotiated case by case. Standard API users should assume the public
$1/M and $5/M rates unless they have a separate enterprise agreement.Q · 08 How accurate is the tokenizer estimate? +
The live counter uses a
3.5 English characters-per-token estimate for Claude-class BPE tokenization. It is useful for planning, but final billing uses Anthropic's server-side token count. Code, tables, and non-Latin scripts can differ materially from the English estimate. Haiku 5.5 uses Anthropic's newer tokenizer, which counts the same text as roughly 30% more tokens than Haiku 4.5.