Anthropic's Claude Haiku 5.5 costs $0.10/$0.50, a tenth of Haiku 4.5

Anthropic released Claude Haiku 5.5 (claude-haiku-5-5) on October 7, 2026. It is the company’s new small model for high-volume work such as classification, routing, extraction and subagent tasks. It runs on the Claude API, Amazon Bedrock, Claude Platform on AWS, Google Cloud and Microsoft Foundry.

What it costs

Haiku 5.5 has two price levels, set by prompt length:

Per 1M tokens	Haiku 5.5, prompts up to 100K	Haiku 5.5, prompts over 100K	Haiku 4.5
Input	$0.10	$0.50	$1.00
Output	$0.50	$2.50	$5.00
Cache write	$0.125	$0.625	$1.25
Cache read	$0.01	$0.05	$0.10

One catch: Haiku 5.5 uses Anthropic’s newer tokenizer, so the same text counts as roughly 30% more tokens than on Haiku 4.5. The per-token cut is still large, but the saving per request is smaller than the table suggests.

What changed from Haiku 4.5

• Context window grows from 200,000 to 1M tokens; maximum output from 64,000 to 128,000 tokens.
• Adaptive thinking is on by default, with an effort setting to trade quality for speed and cost.
• The smallest prompt that can be cached drops from 4,096 to 512 tokens.
• Breaking changes: manual thinking budgets (budget_tokens), non-default temperature, top_p or top_k, and assistant-message prefill now return errors.

How it scores

Anthropic’s published results put Haiku 5.5 far ahead of Haiku 4.5 but behind Sonnet 5.5. On OSWorld 2.1 (offline subset) it scores 72.4%, against 15.7% for Haiku 4.5 and 83.9% for Sonnet 5.5. On Terminal-Bench 4.0 it scores 39.2%, against 70.6% for Sonnet 5.5. Anthropic itself says Sonnet 5.5 and Opus 5.5 remain the better choice for complex agentic coding.

What to do

If you run Haiku 4.5, count your real prompts with the new tokenizer and check which ones pass 100,000 tokens: above that line, Haiku 5.5 costs five times as much per token.