Anthropic has officially unveiledClaude Haiku 4.5, a compact yet powerful addition to its Claude family of language models. The release, which comes just two weeks after the debut of the frontier Claude Sonnet 4.5, brings near-frontier performance to users at a fraction of the cost and at significantly greater speeds.?

Claude Haiku 4.5 is designed to be three times cheaper and more than twice as fast as Sonnet 4, while delivering comparable levels of intelligence, especially in coding, reasoning, and real-time performance tasks.?

Benchmark Results and Performance Highlights

On the SWE-bench Verified coding benchmark, Haiku 4.5 scores 73.3%, rivaling major competitors such as OpenAI’s GPT-5 in debugging and software engineering tests. It also runs hybrid reasoning processes, dynamically adjusting computing power for complex workflows. The model can process 200,000-token multimodal inputs and output up to 64,000 tokens, handling large documents and technical datasets efficiently.?

Anthropic reports that Haiku 4.5 trails the flagship Sonnet 4.5 by less than 10% in most performance benchmarks, while outpacing Sonnet 4 in coding and mathematical reasoning tasks. It achieves this through its extended thinking feature, allowing step-by-step reasoning and multi-agent orchestration for complex problem solving.?

Pricing and Developer Access

Claude Haiku 4.5 is available everywhere today. If you’re a developer, simply use claude-haiku-4-5 via the Claude API. Pricing is now $1/$5 per million input and output tokens.

Working in Tandem with Sonnet 4.5

A key innovation of Anthropic’s ecosystem is interoperability among models. Developers can use Claude Sonnet 4.5 to design multi-step workflows while tasking multiple Haiku 4.5 instances to perform subtasks in parallel — a strategy that increases productivity and lowers infrastructure demands.?

This synergy allows scaling AI applications from rapid prototyping to full-stack deployment with minimal latency.

Haiku 4.5 is now available worldwide on Claude.ai (web, iOS, Android) and through Claude Code. For technical benchmarks, model cards, and deployment documentation, users can refer to Anthropic’s official resources on the Claude platform.