Claude Haiku

Overview

This is a model family overview. For version-specific details, see the individual model entries linked below.

Claude Haiku is the speed-and-cost tier in Anthropic’s lineup. It sacrifices some capability depth compared with Sonnet and Opus but delivers significantly faster responses at a fraction of the price. Haiku is the right choice when throughput, latency, or cost matter more than peak reasoning quality.

Current Latest

Claude Haiku 4.5 remains the efficiency-oriented current reference and the current public Haiku model on Anthropic’s model page. Anthropic’s current model overview lists it with a 200K-token context window, 64K max output, $1 per 1M input tokens, and$5 per 1M output tokens.

Use the full Claude API ID claude-haiku-4-5-20251001 when pinned reproducibility matters; Anthropic also exposes the shorter claude-haiku-4-5 alias. Current docs list extended thinking as supported and adaptive thinking as not supported.

Strengths

• Very fast response times, well suited for real-time applications
• Lowest per-token cost in the Claude family
• Strong coding and computer-use performance for its size and price
• Multimodal support with tool use and structured-output workflows
• Good fit for subagent or parallelized high-volume work

When to Choose Haiku

Choose Haiku when speed or cost is the primary concern:

• High-volume classification, extraction, and routing tasks
• Real-time chat applications where latency matters most
• Coding subagents and multi-agent orchestration where cost matters
• Bulk content processing pipelines
• Cost-sensitive applications that can tolerate occasional quality drops on hard tasks

Access

• Anthropic API as claude-haiku-4-5-20251001, with claude-haiku-4-5 as the Claude API alias
• Amazon Bedrock
• Google Cloud Vertex AI
• Microsoft Foundry
• Claude.ai on web, iOS, and Android
• Claude products and Claude Code