Anthropic pricing
Claude Haiku 4.5 Cost Calculator
Claude Haiku 4.5 — strong everyday performance with aggressive caching discounts (10x cheaper cached input).
Input
$1.00
per 1M tokens
Output
$5.00
per 1M tokens
Context window
200K
tokens
Released
2025-10
Cutoff 2025-02
≈ Estimated tokenizer·$1.00 in·$5.00 out (per 1M)
Quick start with a use case
Total cost per call$0.003500
Input$0.001000
Output$0.002500
Cost comparison
Standard
$0.003500
With Caching
$0.003050
Save 13% ↓
With Batch
$0.001750
Save 50% ↓
Detailed pricing
Claude Haiku 4.5 pricing breakdown
All pricing dimensions including caching and batch discounts.
| Type | Price (per 1M tokens) |
|---|---|
| Input | $1.0000 |
| Output | $5.0000 |
| Cached input | $0.1000 |
| Cache write | $1.2500 |
| Batch input | $0.5000 |
| Batch output | $2.5000 |
Last verified 2026-05-01 · Anthropic official pricing
How it compares
Claude Haiku 4.5 vs alternatives
Single-call cost (1000 input + 500 output tokens) ranked from cheapest.
| Model | Per call |
|---|---|
Claude Haiku 4.5 Anthropic · this page | $0.003500 |
| GPT-5 mini OpenAI | $0.000600 |
| DeepSeek V4 DeepSeek | $0.000900 |
| Gemini 3.0 Flash Google | $0.001250 |
| o4-mini OpenAI | $0.002700 |
| Mistral Large 3 Mistral | $0.006250 |
Recommended use
When to choose Claude Haiku 4.5
Claude Haiku 4.5 shines for general-purpose tasks, high-throughput, and low-latency tasks. Token counts are estimated within ~10-20% margin.
✓Context window of 200K tokens handles long conversations and large documents.
✓Prompt caching available — significant savings for repeated system prompts.
✓Batch API support for non-realtime workloads at ~50% discount.
✓Tool / function calling supported.
FAQ
Frequently asked questions
Claude Haiku 4.5 costs $1.00 per 1M input tokens and $5.00 per 1M output tokens. A typical chat call (1000 input + 500 output tokens) costs approximately $0.0035. Use the calculator above to estimate your specific use case.
Related
Other models you might consider
OpenAI
GPT-5 mini
OpenAI's affordable everyday workhorse
$0.20 in$0.80 out
DeepSeek
DeepSeek V4
DeepSeek's open MoE flagship — top open-weight performer
$0.30 in$1.20 out
Google
Gemini 3.0 Flash
Google's high-throughput multimodal model
$0.25 in$2.00 out
OpenAI
o4-mini
OpenAI's reasoning model — cheaper than o3
$0.90 in$3.60 out
Mistral
Mistral Large 3
Mistral's flagship multilingual open-weight model
$2.50 in$7.50 out