Claude Haiku 5.5 pricing, $0.10 in and $0.50 out per 1M tokens
Claude Haiku 5.5 costs $0.10 per 1M input tokens while your prompt is up to 100,000 tokens, and five times that for prompts over 100,000 tokens. This page gives the official price grid for both bands, the Batch API rates, the cache prices, a calculator for the 100,000-token step and a fair comparison with Claude Haiku 4.5.
Snapshot date: 19 October 2026. Each price on this page is from the Claude pricing page (USD per 1M tokens), fetched that day. Limits and dates are from the Claude release notes, the Claude Haiku 5.5 overview, the Claude Haiku 4.5 overview and the context windows page. Prices can change without notice.
Claude Haiku 5.5 is the budget model in the current Claude line. Its price depends on one boundary: the length of the prompt. A request whose prompt is up to 100,000 tokens pays the lower prices, and a request whose prompt is over 100,000 tokens, even by one token, pays the higher prices for the whole request. Prompt length includes uncached input, cache reads and cache writes. Anthropic applies the price band to each request separately, and cache hits do not exempt a request from the higher band. Source: Claude pricing.
The full official price grid
USD per 1M tokens. Source: Claude pricing, model pricing, snapshot 19 October 2026
| Price line | Prompts up to 100,000 tokens | Prompts over 100,000 tokens |
|---|---|---|
| Base input | $0.10 | $0.50 |
| Cache write, 5-minute | $0.125 | $0.625 |
| Cache write, 1-hour | $0.20 | $1.00 |
| Cache hit (read) | $0.01 | $0.05 |
| Output | $0.50 | $2.50 |
Batch API prices
Batch API rates, USD per 1M tokens. Source: Claude pricing, batch processing, snapshot 19 October 2026
| Batch price line | Prompts up to 100,000 tokens | Prompts over 100,000 tokens |
|---|---|---|
| Batch input | $0.05 | $0.25 |
| Batch output | $0.25 | $1.25 |
US-only inference: the Claude pricing page says that for Claude 4.6 and later models, setting inference_geo: "us" multiplies every token price (input, output, cache writes and cache reads) by 1.1. On Haiku 5.5, this also applies to the prices for prompts over 100,000 tokens. This applies on the Claude API and Claude Platform on AWS, and to US Data Zone Standard deployments in Microsoft Foundry. Bedrock and Google Cloud have their own regional pricing. Haiku 4.5 is an earlier model: requests that include inference_geo return a 400 error. Source: Claude pricing, data residency, snapshot 19 October 2026.
How the 100,000-token step works
One boundary, one price for the whole request
Three rules decide which band prices a request pays:
- The boundary is the prompt length: a request whose prompt is up to 100,000 tokens pays the lower prices, and a request whose prompt is over 100,000 tokens, even by one token, pays the higher prices for the whole request.
- The prompt length counts all input tokens, including cache reads and cache writes.
- Requests are priced one by one, and earlier requests keep the prices they were billed at.
Example (KickLLM calculation, official rates of the snapshot date): a request with a 120,000-token prompt and 4,000 output tokens pays the over-100,000 band on every token: 120,000 × $0.50 / 1,000,000 + 4,000 × $2.50 / 1,000,000 = $0.06 + $0.01 = $0.07. If the same prompt had counted 99,000 tokens, the bill would be 99,000 × $0.10 / 1,000,000 + 4,000 × $0.50 / 1,000,000 = $0.0119, so on this request shape the bill is about 5.9 times higher ($0.07 against $0.0119). The per-token rates in the over-100,000 band are five times the rates of the up-to band; the rest of the difference comes from the longer prompt.
Haiku 5.5 cost calculator
KickLLM estimate. USD per 1M tokens, standard API. Prices: Claude pricing; limits: Haiku 5.5 overview and Haiku 4.5 overview. Snapshot 19 October 2026
KickLLM estimate, not an Anthropic figure. Prompt length = uncached input + cache reads + cache writes, as the pricing page defines it. Documented per-request limits used: 1,000,000-token context window (prompt + output) and 128,000 max output tokens for Haiku 5.5; 200,000 and 64,000 for Haiku 4.5. The Batch option also uses the 128,000 limit; the 300,000-token Batch output with the output-300k-2026-03-24 beta header is not modeled. Batch cache prices are derived (cache multiplier × 50% batch discount); the official batch table lists only input and output. US-only with Batch is also a KickLLM derivation (1.1x applied to the batch rates). Sources: Claude pricing, Claude context windows, fetched 19 October 2026.
Claude Haiku 5.5 vs Claude Haiku 4.5
From the pricing page, the two overview pages and What's new in Claude Haiku 5.5, snapshot 19 October 2026
| Model | Base input | 5m cache write | 1h cache write | Cache hit | Output | Batch in / out | Context window | Max output | Min cacheable prompt |
|---|---|---|---|---|---|---|---|---|---|
| Claude Haiku 5.5 (prompt up to 100,000 tokens) | $0.10 | $0.125 | $0.20 | $0.01 | $0.50 | $0.05 / $0.25 | 1M | 128K | 512 tokens |
| Claude Haiku 5.5 (prompt over 100,000 tokens) | $0.50 | $0.625 | $1.00 | $0.05 | $2.50 | $0.25 / $1.25 | 1M | 128K | 512 tokens |
| Claude Haiku 4.5 (flat) | $1.00 | $1.25 | $2.00 | $0.10 | $5.00 | $0.50 / $2.50 | 200K | 64K | 4,096 tokens |
For prompts up to 100,000 tokens the list price is one tenth of Haiku 4.5. With about 30% more tokens for the same text, the same text costs about 0.13 times the Haiku 4.5 cost (0.1 × 1.3), roughly one eighth. This is approximate: Anthropic says the exact increase depends on the content. For prompts over 100,000 tokens it is one half of the Haiku 4.5 price (Haiku 4.5 has one flat price). Haiku 5.5 was released on October 7, 2026, according to the Claude release notes and the Haiku 5.5 overview. The Claude Haiku 5.5 overview gives its retirement as "not sooner than October 7, 2027". The Claude Haiku 4.5 overview lists Haiku 4.5 as "Active (legacy)" with retirement "not sooner than October 15, 2026".
The Batch API gives a 50% discount on input and output tokens. Haiku 5.5 supports up to 300k output tokens on the Batch API with the output-300k-2026-03-24 beta header, while the synchronous Messages API limit is 128k. The minimum cacheable prompt drops from 4,096 tokens on Haiku 4.5 to 512 tokens on Haiku 5.5, which matters for small cached prompts. The Haiku 4.5 prices come from the same pricing page; its 200K context window and 64K max output come from the Claude Haiku 4.5 overview.
How these numbers were checked
There are two kinds of sources, all fetched as text on 19 October 2026: the Claude pricing page, which gives both prompt bands, the cache write and hit lines, the Batch prices, the data residency multiplier and the rules for how the prompt length is counted; and the model overview pages, Claude Haiku 5.5 overview and Claude Haiku 4.5 overview, which give the context windows, max output and the retirement dates. The minimum cacheable prompts (512 and 4,096 tokens) and the note that the exact token increase depends on the content are from What's new in Claude Haiku 5.5.
The calculator outputs are KickLLM estimates. They are not Anthropic figures, and your invoice can be different. KickLLM is not affiliated with Anthropic or any other model provider.
