Claude Haiku 4.5
anthropic/claude-haiku-4.5Claude Haiku 4.5 is Anthropic’s fastest and most efficient model, delivering near-frontier intelligence at a fraction of the cost and latency of larger Claude models. Matching Claude Sonnet 4’s performance across reasoning, coding, and computer-use tasks, Haiku 4.5 brings frontier-level capability to real-time and high-volume applications.
- Input price
- $1.00USD / 1M tokens
- Output price
- $5.00USD / 1M tokens
- Cache read
- $0.10USD / 1M tokens
- Cache write
- $1.25USD / 1M tokens
- Context window
- 200K
- Max output
- 64K
Specifications
- Input
- Text, image
- Output
- Text
About this model
Claude Haiku 4.5 is the fast, low-cost tier of Anthropic's Claude 4.5 family. It reads text and images, writes text, and has a 200K-token context window. Anthropic added extended thinking to Haiku with this release; you turn it on per request through the Messages API.
When to use it
- Real-time chat and support assistants where latency matters.
- Sub-agents that handle many small steps for a larger model.
- Coding help and code review at high volume.
Call it
The model works on Chat Completions (/v1/chat/completions) and on Anthropic's Messages API (/v1/messages). Use Messages when you want extended thinking.
from openai import OpenAI
client = OpenAI(base_url="https://api.sandbase.ai/v1", api_key="$SANDBASE_API_KEY")
resp = client.chat.completions.create(
model="anthropic/claude-haiku-4.5",
messages=[{"role": "user", "content": "Explain what an API gateway does."}],
max_tokens=512,
)
print(resp.choices[0].message.content)
What works on SandBase
Checked with real calls through SandBase on October 28, 2026.
| Feature | Works | Notes |
|---|---|---|
| Streaming | Yes | stream: true, with usage in the last chunk when stream_options.include_usage is set. |
| Tool calling | Yes | tools and tool_choice; returns tool_calls. |
| Image input | Yes | image_url parts with an https URL or a data: URI. |
| Extended thinking | Messages API | Send thinking: {"type": "enabled", "budget_tokens": 2000} to /v1/messages; the reply starts with a thinking block. reasoning_effort on Chat Completions has no effect. |
| Structured output | No | response_format on Chat Completions and output_format on Messages are both ignored; the model answers in prose. Describe the JSON you want in the prompt, or define a tool whose input schema is the shape you need. |
Limits and parameters
- Context window: 200,000 tokens.
- Maximum output: 64,000 tokens per request (
max_tokens). - Accepts
temperature,top_pandstopon Chat Completions, andstop_sequenceson Messages.
Example response
Real response to "In two sentences, explain what an API gateway does." through Chat Completions, trimmed.
{
"id": "431cc58b-9b0…",
"object": "chat.completion",
"model": "anthropic/claude-haiku-4.5",
"choices": [
{
"index": 0,
"finish_reason": "stop",
"message": {
"role": "assistant",
"content": "An API gateway is the single entry point that sits in front of a set of backend services and routes each incoming client request to the service that handles…"
}
}
],
"usage": {
"completion_tokens": 98,
"completion_tokens_details": {
"reasoning_tokens": 0
},
"prompt_tokens": 18,
"prompt_tokens_details": {
"cache_write_tokens": 0,
"cached_tokens": 0
},
"total_tokens": 116
}
}
Pricing
| Tokens | Price per 1M |
|---|---|
| Input | $1.00 |
| Output | $5.00 |
| Cached input (read) | $0.10 |
| Cache write, 5-minute | $1.25 |
| Cache write, 1-hour | $2.00 |
Prompt caching works through cache_control blocks on /v1/messages; a repeat call reports the cached prefix as cache_read_input_tokens. Reads cost 10% of the input price.
Related models
- Claude Haiku 5.5: the current Haiku tier.
- Claude Sonnet 4.5: the Sonnet model of the same generation, for harder tasks.
Try the model
Playground
Write a prompt to try this model. Usage is billed at the prices shown.
Try a prompt
Add an image (optional)
Your response will appear here
Choose an example or write a prompt, then click Run.
Start building
Send your first request
Choose either supported API with unified authentication and usage tracking.
# The quoted heredoc keeps Unicode and shell metacharacters unchanged.
result=$(curl --fail-with-body --silent \
-X POST "https://api.sandbase.ai/v1/chat/completions" \
-H "Authorization: Bearer $SANDBASE_API_KEY" \
-H "Content-Type: application/json" \
--data-binary @- <<'SANDBASE_JSON'
{
"model": "anthropic/claude-haiku-4.5",
"messages": [
{
"role": "user",
"content": "Hello"
}
],
"max_tokens": 2048
}
SANDBASE_JSON
)
printf '%s\n' "$result"Input Schema
10 parameters · 1 required · 9 optional
| Parameter | Type | Required | Description |
|---|---|---|---|
messages | object[] | Required | — |
stop | string[] | Optional | — |
tools | array | Optional | — |
top_k | integer | Optional | Min: 0 |
top_p | number | Optional | Min: 0 · Max: 1 |
stream | boolean | Optional | Default: false |
max_tokens | integer | Optional | Max: 64000 |
temperature | number | Optional | Min: 0 · Max: 2 · Default: 1 |
tool_choice | string | Optional | — |
response_format | object | Optional | — |
Output Schema
| Field | Type | Description |
|---|---|---|
id | string | Unique identifier for the completion |
object | string | Object type, e.g. "chat.completion" |
model | string | Model used for the completion |
choices | array | List of completion choices |
choices[].message.content | string | Generated text content |
choices[].finish_reason | string | Reason the generation stopped |
usage.prompt_tokens | integer | Number of tokens in the prompt |
usage.completion_tokens | integer | Number of tokens in the completion |
usage.total_tokens | integer | Total tokens used |
Questions
FAQ
How do I call Anthropic Claude Haiku 4.5 through SandBase?
Create a SandBase API key, then send requests with the model ID "anthropic/claude-haiku-4.5" to Chat Completions API (/v1/chat/completions) or Messages API (/v1/messages) at https://api.sandbase.ai. The request examples on this page show the exact payload.
How much does Anthropic Claude Haiku 4.5 cost?
$1 per 1M input tokens and $5 per 1M output tokens, billed per request from your SandBase balance. Discounts, when available, are shown in the pricing on this page.
What is the context window of Anthropic Claude Haiku 4.5?
200K tokens of context, with up to 64K output tokens per response.
Which features does Anthropic Claude Haiku 4.5 support?
Anthropic Claude Haiku 4.5 supports vision (image input), reasoning, function calling, structured output. See Specifications above for the full list.
Do I need a separate Anthropic account?
No. One SandBase API key and balance gives you access to Anthropic Claude Haiku 4.5 and the other models in the catalog; you do not need to sign up with Anthropic separately.
