Claude Haiku 4.5
Fast and efficient for high-volume, low-latency work.
Haiku 4.5 is tuned for speed and throughput — the model to reach for when you're running lots of requests and want snappy responses at a low price.
- High volume
- Low latency
- Classification
Model details
- Context window
- 200K tokens
- Max output
- 64K tokens
- Input
- Text, images
- Output
- Text
- Capabilities
- Reasoning, Vision, Tool calling, Streaming, JSON output
- Interfaces
- OpenAI-style Chat Completions, Anthropic-style Messages, OpenAI-style Responses
Pricing
Prices use the units shown below. No gateway fees. Usage, size and model options may affect the total.
| Metric | Price |
|---|---|
| / 1M input tokens | $1 |
| / 1M output tokens | $5 |
| / 1M cached input tokens | $0.1 |
| / 1M cache write tokens (5 min) | $1.25 |
| / 1M cache write tokens (1 hour) | $2 |
| / web search | $0.01 |
Why teams pick it
- Fast responses for latency-sensitive flows
- Priced for high-throughput workloads
- Vision and tool calling included
Call Claude Haiku 4.5
Claude Haiku 4.5 answers on any of the APIs below — pick the one your code already speaks and just swap in your Cloakray key.
/v1/chat/completionsOpenAI-style Chat Completions · openai SDK or plain HTTP
curl https://api.cloakray.com/v1/chat/completions \
-H "Authorization: Bearer ckr_..." \
-H "Content-Type: application/json" \
-d '{
"model": "claude-haiku-4-5",
"messages": [
{
"role": "user",
"content": "Hello"
}
]
}'/v1/messagesAnthropic-style Messages · @anthropic-ai/sdk SDK or plain HTTP
curl https://api.cloakray.com/v1/messages \
-H "Authorization: Bearer ckr_..." \
-H "Content-Type: application/json" \
-d '{
"model": "claude-haiku-4-5",
"max_tokens": 1024,
"messages": [
{
"role": "user",
"content": "Hello"
}
]
}'/v1/responsesOpenAI-style Responses · openai SDK or plain HTTP
curl https://api.cloakray.com/v1/responses \
-H "Authorization: Bearer ckr_..." \
-H "Content-Type: application/json" \
-d '{
"model": "claude-haiku-4-5",
"input": "Hello"
}'Conversation compaction is also available at POST /v1/responses/compact.
More from Anthropic
Claude Fable 5.1Anthropic's most intelligent model, for the hardest problems.$10 / 1M input tokens · $50 / 1M output tokens · See all rates on the model pageClaude Opus 5.5Long-running agentic coding at a lower price than Opus 5.$4 / 1M input tokens · $20 / 1M output tokens · See all rates on the model pageClaude Fable 5Next-generation intelligence for knowledge work and coding.$10 / 1M input tokens · $50 / 1M output tokens · See all rates on the model page