Anthropic API Pricing for Claude Haiku 5.5: $0.10 Input, $0.50 Output per Million Tokens
Claude Haiku 5.5 costs $0.10 per million input tokens and $0.50 per million output tokens when the prompt is 100,000 tokens or shorter. Above that size the rates are $0.50 and $2.50. Anthropic released the model on October 7, 2026, and the prices below are as of October 15, 2026, from Anthropic's Haiku 5.5 page and the Claude pricing page.
The table
All figures are USD per million tokens, from Anthropic's Haiku 5.5 page and the Claude pricing page. The model ID is claude-haiku-5-5.
| Charge | Haiku 5.5, prompt up to 100K | Haiku 5.5, prompt over 100K | Haiku 4.5 |
|---|---|---|---|
| Input | $0.10 | $0.50 | $1.00 |
| Output | $0.50 | $2.50 | $5.00 |
| Cache read | $0.01 | $0.05 | $0.10 |
| Cache write (5 min) | $0.125 | $0.625 | $1.25 |
Source: Anthropic, Haiku 5.5 page and Claude pricing page, checked October 15, 2026.
Anthropic's own wording: up to the threshold Haiku 5.5 costs 90% less than Haiku 4.5, above it 50% less. The pages do not say how a prompt of exactly 100,000 tokens is billed, so keep real requests clear of the line if you can.
Two more lines on the pricing page matter for a budget. Haiku 5.5 with US-only inference is priced at 1.1x for input and output. And the page says batch processing saves 50%, but it lists no separate batch rates for Haiku, so check the page before you put a batch discount into a spreadsheet.
One request, priced
Take a request with 20,000 input tokens and 1,000 output tokens, no caching, at the rates in the table above (Anthropic pricing, October 15, 2026).
- Input: 20,000 / 1,000,000 x $0.10 = $0.0020
- Output: 1,000 / 1,000,000 x $0.50 = $0.0005
- Total: $0.0025
The same request on Haiku 4.5 comes to $0.020 for input plus $0.005 for output, $0.025 in all. Ten times more.
The tokenizer catch
That tenfold gap assumes both models count the text the same way. They do not. Anthropic says Haiku 5.5 uses an updated tokenizer that takes "slightly more tokens per task", and that its figure of about 75% lower average cost against Haiku 4.5 already includes this. The company gives no per-text multiplier, so any number here is an assumption.
Say, purely for illustration, that your text becomes 20% longer in tokens: 24,000 in, 1,200 out. The bill is $0.0024 plus $0.0006, so $0.0030. Still about 88% below $0.025. The rate cut is large enough to absorb a tokenizer change of that size. Your own prompts are the only real test, and a character count will not tell you the token count.
Against April 2023
The calculator on this site's home page still uses the April 2023 rates: GPT-3.5 Turbo at $0.002 per 1K tokens, GPT-4 (8K) at $0.03 per 1K. In per-million terms that is $2 and $30. Haiku 5.5's input rate of $0.10 is one twentieth of the first and one three-hundredth of the second. The comparison is only about price. The models are not the same kind of thing, and the 2023 calculator shows a single rate per model, with no separate output price.
Prices move, so treat the table as a dated snapshot. The two Anthropic pages are the reference; the calculator here is for rough numbers.