Claude Haiku 4.5 API price
Anthropic lists Claude Haiku 4.5 at $1 per million input tokens and $5 per million output tokens on the Claude API. Those rates are in USD; Anthropic’s model page labels Haiku 4.5 a legacy model and identifies Haiku 5.5 as the current Haiku model. (platform.claude.com)
| Usage | Listed price |
|---|---|
| Input tokens | $1.00 / million tokens |
| Output tokens | $5.00 / million tokens |
| 5-minute cache write | $1.25 / million tokens |
| 1-hour cache write | $2.00 / million tokens |
| Cache read | $0.10 / million tokens |
| Batch API | 50% discount on input and output |
These are Anthropic’s listed Claude API rates. A cloud provider, gateway, or router may have its own pricing or terms, so don’t assume its final bill will match direct API billing. (platform.claude.com)
Estimate your Haiku 4.5 API cost
Calculate standard token charges this way:
(input tokens ÷ 1,000,000 × $1) + (output tokens ÷ 1,000,000 × $5)
For example, 10 million input tokens and 2 million output tokens would cost $20 at the listed standard rates:
- Input: 10 × $1 = $10
- Output: 2 × $5 = $10
- Total: $20
This estimate covers only the input and output tokens in the example. If your requests use prompt caching or Batch API, the applicable rates differ; Anthropic lists cache prices above and a 50% Batch API discount on input and output. (platform.claude.com)
Model ID and current model status
For the direct Claude API, Anthropic lists the model ID claude-haiku-4-5-20251001 and the alias claude-haiku-4-5. Its documentation describes Haiku 4.5 as a legacy model, with a 200,000-token context window and a maximum output of 64,000 tokens. Anthropic identifies Haiku 5.5 as the current Haiku model, so compare the two if you’re choosing a model for a new deployment; their prices and limits differ. (platform.claude.com)
For production budgeting, estimate with your expected input and output token counts, then test representative requests for quality, latency, and total cost. A lower per-token rate alone doesn’t show which model is the better fit for your workload.
Sources
Claude Haiku 4.5 — Claude Platform Docs