Claude Haiku 4.5: Fast, Cheap AI Model Matching Sonnet 4 Performance

Anthropic announced on October 15 that its newest Claude Haiku 4.5 model is now available on Claude.ai, the Anthropic API, Amazon Bedrock and Google Vertex AI.

Haiku 4.5 is the smallest, fastest and most cost‑effective model in the Claude family, yet its performance rivals the larger Claude Sonnet 4, making it attractive for multi‑agent applications where Sonnet delegates tasks to Haiku‑based sub‑agents.

The model is priced at $1 per million input tokens and $5 per million output tokens, compared with $0.80/$4 for the previous Haiku 3.5.

Benchmark Data

In Anthropic’s internal benchmarks Haiku 4.5 matches or exceeds Sonnet 4 across most metrics and is competitive with Google’s Gemini 2.5 and OpenAI’s GPT‑5, especially in coding, tool use and software‑engineering (SWE) benchmarks, where it sometimes outperforms Sonnet 4.

Haiku 4.5 is a hybrid‑reasoning model that introduces an optional “extended thinking” mode, a first for the Haiku series.

The default context window is 200 k tokens, but developers on the Claude platform can request up to 1 M tokens.

Why the Delay?

According to Anthropic’s developer‑relations lead Alex Albert, the company spent the previous year refining its frontier models. With the launch of Sonnet 4.5, Anthropic now aims to unlock more use cases and align applications with cutting‑edge AI capabilities.

Haiku Multi‑Agent System

Anthropic envisions Haiku 4.5 powering fast, task‑specific sub‑agents in multi‑agent systems. The upcoming Claude Code agent will leverage this capability, while Sonnet 4.5 provides high‑level guidance and Haiku 4.5 executes the tasks, excelling at tool calls.

Jeff Want, CEO of Windsurf, noted that Haiku 4.5 blurs the traditional trade‑off between quality, speed and cost, offering a “fast frontier model” that remains cost‑effective.