Activefastllm Proprietary

Claude 3 Haiku

by Anthropic· Released March 2024· Cutoff August 2023

Claude 3 Haiku is Anthropic's fastest and most compact model in the Claude 3 family, designed for near-instant responsiveness and high throughput. It excels at lightweight tasks like customer service, content moderation, and real-time chat, offering a balance of speed and intelligence at a lower cost.

Official Site API Docs

Input cost

$0.25 per 1M tokens

Output cost

$1.25 per 1M tokens

Context window

200K tokens

Max output

4096 tokens

Modalities

textimage

License

proprietary

Capabilities

Text GenerationCode GenerationFunction CallingStreamingJSON ModeVision (image input)

Best For

Fast, cost-effective responses for high-traffic applications and real-time interactions.

Strengths

Fastest inference speed in Claude 3 family
Lowest cost per token
200K token context window
Vision capabilities for image analysis
Strong instruction following

Limitations

Less capable in complex reasoning compared to Sonnet and Opus
Smaller knowledge base and creativity
Not suitable for advanced coding or nuanced tasks

Use Cases

Customer support chatbots

Content moderation and filtering

Real-time language translation

Quick data extraction and classification

Lightweight code generation and debugging

Image captioning and analysis

Personal assistant applications

Improvements Over Previous Model

First model in the Claude 3 Haiku line, no direct predecessor; compared to Claude 2.1: ~3× faster inference
Context window increased from 100K to 200K tokens
Added vision/image input support
Lower pricing: $0.25/M input tokens vs $8.00/M for Claude 2.1
Improved instruction following and reduced refusals

Back to all models