Claude Haiku 5.5 API price, and what each result really costs
Anthropic, listed 7 Oct 2026. Price checked 12 Nov 2026; unchanged since we started tracking it. Source: Anthropic's pricing page. anthropic/claude-haiku-5.5
List price per million tokens
- Input
- $0.10
- Cached input
- $0.01
- Output
- $0.50
- Context window
- 1M tokens
Note: Price applies to prompts up to 100K tokens; prompts over 100K cost $0.50 in / $0.05 cached / $2.50 out.
What it costs per task, and per task that works
Six common task shapes with typical token counts. The real cost counts the tries that get redone; the rates are examples, not measurements.
| Task | Per 1,000 tries | Per 1,000 that work, 8 in 10 | Per 1,000 that work, 6 in 10 | Per 1,000 that work, 4 in 10 |
|---|---|---|---|---|
| Support bot answer3,000 in / 300 out | $0.45 | $0.56 | $0.75 | $1.13 |
| Invoice field extraction2,500 in / 250 out | $0.38 | $0.47 | $0.63 | $0.94 |
| Lead scoring1,200 in / 100 out | $0.17 | $0.21 | $0.28 | $0.42 |
| Email reply draft1,500 in / 400 out | $0.35 | $0.44 | $0.58 | $0.87 |
| Translation900 in / 800 out | $0.49 | $0.61 | $0.82 | $1.22 |
| Search re-ranking4,000 in / 50 out | $0.42 | $0.53 | $0.71 | $1.06 |
Work out your own real cost per resultTest Claude Haiku 5.5 on your own prompt
Compare Claude Haiku 5.5 with
- Claude Haiku 4.5$1.00 in / $5.00 outSwitch cost
- Claude Sonnet 5.5$2.00 in / $10.00 outSwitch cost
- GPT-6 Luna$0.10 in / $0.50 outSwitch cost
- Gemini 2.5 Flash-Lite$0.10 in / $0.40 outSwitch cost
- Qwen3.8 Flash$0.15 in / $0.47 outSwitch cost
Worked-through switches: GPT-4o → Claude Haiku 5.5: cost and quality on real tasks, GPT-4o mini → Claude Haiku 5.5: cost and quality on real tasks, Claude Sonnet 5.5 → Claude Haiku 5.5: cost and quality on real tasks, Claude Haiku 4.5 → Claude Haiku 5.5: cost and quality on real tasks
Questions
How much does the Claude Haiku 5.5 API cost?
$0.10 per million input tokens and $0.50 per million output tokens, or $0.01 per million for cached input. Price checked 12 Nov 2026; unchanged since we started tracking it.
What does a result that works really cost with Claude Haiku 5.5?
A support bot answer (about 3,000 tokens in, 300 out) costs $0.00045 per try. If 6 in 10 work first time, each answer that works really costs $0.00075, because the redone ones were paid for too.