Claude 4.5 Haiku API

Claude 4.5 Haiku API is Anthropic's fastest and most cost-efficient Claude 4 model. It matches Sonnet 4 performance on coding, computer use, and agent tasks and scores 73.3% on SWE-bench Verified. On EvoLink, you can access it with the documented model enum.

Using coding CLIs? Run Claude 4.5 Haiku via EvoCode — One API for Code Agents & CLIs. (View Docs)

PRICING

PLAN	CONTEXT WINDOW	MAX OUTPUT	INPUT	OUTPUT	CACHE WRITE	CACHE READ
Claude Haiku 4.5	200.0K	64.0K	$0.850-15% $1.00Official Price	$4.25-15% $5.00Official Price	$1.06-15% $1.25Official Price	$0.086-14% $0.100Official Price
Claude Haiku 4.5 (Beta)	200.0K	64.0K	$0.260-74% $1.00Official Price	$1.30-74% $5.00Official Price	$0.325-74% $1.25Official Price	$0.026-74% $0.100Official Price

Web Search Tool

Server-side web search capability

$0.011/search

Pricing Note: Price unit: USD / 1M tokens

Cache Hit: Price applies to cached prompt tokens.

Two ways to run Claude Haiku 4.5 — pick the tier that matches your workload.

· Claude Haiku 4.5: the default tier for production reliability and predictable availability.
· Claude Haiku 4.5 (Beta): a lower-cost tier with best-effort availability; retries recommended for retry-tolerant workloads.

Claude 4.5 Haiku API for Fast, Scaled AI

Anthropic positions Haiku 4.5 for parallelized execution, sub-agents, and high-volume operations, as well as latency-sensitive experiences and free-tier scale. Choose Claude 4.5 Haiku API when you need speed-first performance without the cost profile of larger models.

Hero showcase of Claude 4.5 Haiku API feature 1

What can you build with the Claude 4.5 Haiku API?

Free-Tier and High-Scale AI

Claude 4.5 Haiku API is positioned to make AI practical at scale. It excels at parallelized execution and high-volume operations, enabling free-tier assistants, onboarding helpers, and large workflows where cost and response speed drive product adoption.

Start building

Use-case showcase of Claude 4.5 Haiku API feature 2

Latency-Sensitive Support

Anthropic highlights Haiku 4.5 for real-time applications like customer service agents and chatbots. That makes it a strong choice for fast support flows, ticket triage, and live help experiences where response time shapes user satisfaction.

Optimize support

Use-case showcase of Claude 4.5 Haiku API feature 3

Coding Sub-Agents

Claude 4.5 Haiku API matches Sonnet 4 performance on coding tasks, enabling coding sub-agents that tackle refactors, migrations, and large feature builds with quality and speed. Use it to accelerate developer workflows while keeping quality aligned with higher-tier models.

Explore coding

Use-case showcase of Claude 4.5 Haiku API feature 4

Why teams choose the Claude 4.5 Haiku API

Claude 4.5 Haiku API combines near-frontier coding quality with speed and efficiency for scaled, user-facing products.

Fastest Claude 4 Model

Anthropic describes Haiku 4.5 as the fastest and most cost-efficient in Claude 4.

Near-Frontier Coding

Matches Sonnet 4 on coding and scores 73.3% on SWE-bench Verified.

High-Volume Operations

Optimized for parallelized execution, sub-agents, and large request volumes.

How to integrate the Claude 4.5 Haiku API

Use the EvoLink API with the documented model enum and required fields, then tune optional controls as needed.

Step 1 — Authenticate

Create an EvoLink API key and send requests with Bearer token authentication.

Step 2 — Set required fields

Provide model: claude-haiku-4-5-20251001, max_tokens, and a messages array.

Step 3 — Tune outputs

Adjust system, temperature, top_p, top_k, stop_sequences, stream, tools, and tool_choice for your use case.

View API Docs

Core Claude 4.5 Haiku API capabilities

Model facts from Anthropic, plus EvoLink access details

Positioning

Fastest and Most Cost-Efficient

Anthropic positions Claude 4.5 Haiku as the fastest and most cost-efficient model in the Claude 4 family, optimized for speed-first deployments.

Benchmarks

73.3% SWE-bench Verified

Claude 4.5 Haiku matches Sonnet 4 performance on coding, computer use, and agent tasks and scores 73.3% on SWE-bench Verified.

Scale

Parallelized Execution

Haiku 4.5 excels at parallelized execution, sub-agents, and high-volume operations for large-scale workloads.

Latency

Latency-Sensitive Experiences

Designed for real-time applications like customer service agents and chatbots where response time is critical.

Agents

Coding Sub-Agents

Positioned for sub-agents that handle refactors, migrations, and large feature builds with speed and quality.

Availability

Availability Across Platforms

Haiku 4.5 is available on Claude.ai, the Claude Developer Platform, Amazon Bedrock, Google Cloud Vertex AI, Microsoft Foundry, and Claude Code.

Claude 4.5 Haiku API FAQs

Everything you need to know about the product and billing.

Anthropic describes Claude 4.5 Haiku as the fastest and most cost-efficient model in the Claude 4 family. It matches Sonnet 4 performance on coding, computer use, and agent tasks and scores 73.3% on SWE-bench Verified. These facts position it as a speed-first model that still delivers near-frontier quality for real-world applications.

Anthropic highlights parallelized execution, sub-agents, and high-volume operations. Use cases include free-tier products and budget-conscious applications, latency-sensitive customer service agents and chatbots, coding sub-agents for refactors and migrations, plus financial analysis and research sub-agents that process many sources at once.

Anthropic states Haiku 4.5 matches Sonnet 4 performance on coding tasks, computer use, and agent tasks. It also reports a 73.3% score on SWE-bench Verified, indicating strong coding capability despite its speed-first profile.

Anthropic lists availability on Claude.ai (web, iOS, Android), the Claude Developer Platform, Amazon Bedrock, Google Cloud Vertex AI, Microsoft Foundry, and Claude Code. This gives teams flexibility across direct API access, cloud marketplaces, and coding workflows.

Yes. Anthropic states Claude Haiku 4.5 is available in Claude Code, which is useful for teams that want fast coding assistance in CLI or developer workflows.

Use the model enum `claude-haiku-4-5-20251001` in the request body. The EvoLink Claude Messages API lists model IDs explicitly, and Haiku 4.5 is mapped to that enum. Using the documented model ID ensures the request routes to the correct model.

The EvoLink Claude Messages API requires `model`, `max_tokens`, and `messages`. Each message includes a `role` and `content`, and the API expects alternating user and assistant turns. This structure keeps requests predictable and ensures the model can generate the next assistant response correctly.

You can pass optional parameters such as `system`, `temperature`, `top_p`, `top_k`, and `stop_sequences`. The API also supports `stream` and accepts `tools` with `tool_choice`. These controls help you tune style, randomness, and tool usage behavior for Claude 4.5 Haiku API responses.