Claude Haiku 5.5

Available

An Anthropic model with a 1M-token context window, image input and tool calling. On the Anthropic API since October 7, 2026.

Access details: Anthropic API

The model ID to use in API requests.

claude-haiku-5-5

Updated Oct 20, 2026

Sources & notes

Capabilities & limits

Context window: 1M
Maximum output: 128K
Image input: Supported
Tool calling: Supported
Structured output: Supported

Pricing

USD / 1M tokens

Input: $0.10
Output: $0.50

Standard rates per million tokens. Discounts and long-context rates, where offered, are on Anthropic's pricing page.

Full rate card 10 rates

All prices in USD.

Input tokens

$0.10 per 1M tokens

prompts up to 100,000 tokens

Output tokens

$0.50 per 1M tokens

prompts up to 100,000 tokens

Long-prompt input tokens

$0.50 per 1M tokens

prompts over 100,000 tokens

Long-prompt output tokens

$2.50 per 1M tokens

prompts over 100,000 tokens

Cache reads

$0.01 per 1M tokens

prompt caching
prompts up to 100,000 tokens

5 minute cache writes

$0.125 per 1M tokens

prompt caching
5 minute duration
prompts up to 100,000 tokens

1 hour cache writes

$0.20 per 1M tokens

prompt caching
1 hour duration
prompts up to 100,000 tokens

Long-prompt cache reads

$0.05 per 1M tokens

prompt caching
prompts over 100,000 tokens

Long-prompt 5 minute cache writes

$0.625 per 1M tokens

prompt caching
5 minute duration
prompts over 100,000 tokens

Long-prompt 1 hour cache writes

$1.00 per 1M tokens

prompt caching
1 hour duration
prompts over 100,000 tokens

Priced by prompt length: prompts over 100,000 tokens pay the higher rates. Batch API requests are discounted 50% on input and output.

Price history

Oct 20, 2026

Input $0.10 per million tokens

Oct 20, 2026

Output $0.50 per million tokens

Technical details

Parameters: Not published
Active parameters: Not published
Architecture: Not published
Training / knowledge cutoff: Jun 2026
Input modalities: Text, image
Output modalities: Text

From Anthropic

Official announcement ↗

How they describe it

For high-volume, latency-sensitive tasks such as classification, extraction, and routing

What they say is new

the cheapest, fastest, and most capable small model we've ever released.

Release & lifecycle

First available: Oct 7, 2026
available: As of Oct 20, 2026