Model ComparisonAnthropic

Claude Haiku 3.5

Claude 3.5 Haiku — the cheapest Anthropic model for high-volume classification and extraction.

VisionTool usePrompt cachingReasoningFast
Pricing (verified 2026-08-19)
TypePrice / 1M tokensNotes
Input$0.80Per 1M input tokens
Output$4.00Per 1M output tokens
Cache read$0.080Cached prefix read-back
Cache write$1.00Anthropic 5m TTL; 2× for 1h TTL
Source: Anthropic official pricing page · Last verified 2026-08-19
Context Window
200K tokens
150K words · 1 pages of text

At $0.80/M input, Claude 3.5 Haiku is Anthropic's answer to GPT-5 mini. Smarter than you'd expect, and a natural fit for the classification, tagging, and extraction layer of any LLM pipeline.

Best for

  • High-volume classification and tagging
  • Data extraction from documents
  • Simple Q&A chatbots with light reasoning
  • Triage layer before routing to expensive models

Not ideal for

  • Complex reasoning tasks
  • Long creative writing
  • Tasks requiring maximum Anthropic model quality

Real cost scenarios

ScenarioEst. monthly costBreakdown
Email classification — 500k emails/month$200250M tokens at $0.80/M = $200
Data extraction — 1M records/month$400500M tokens at $0.80/M = $400

Estimates with typical token distributions. Use the Prompt Cost Calculator for your exact workload.

How to cut costs on Claude Haiku 3.5

  1. Pair Haiku 3.5 with Claude Sonnet 5 in a router: Haiku handles simple requests, Sonnet escalates for complex ones. Typical cost reduction: 60-70%.

Calculate your Claude Haiku 3.5 costs

Prompt Cost Calculator
Exact cost for your prompt
Agent Cost Calculator
Multi-step agent loops
Prompt Caching Calculator
Caching savings
Conversation Cost
Chat history strategies

Other Anthropic models

Claude Opus 5$5.00/MClaude Sonnet 5$3.00/MClaude Haiku 4.5$1.00/M

Compare Claude Haiku 3.5

gpt 5 mini vs claude haiku 4 5gemini 3 flash vs claude haiku 4 5

Frequently Asked Questions

How does Claude 3.5 Haiku compare to Claude Haiku 4.5?

Haiku 4.5 ($1.00/M input) is 25% more expensive but noticeably smarter. For most classification tasks, 3.5 Haiku is sufficient. For tasks requiring better reasoning or coding, upgrade to Haiku 4.5 or Sonnet 5.

Pricing verified 2026-08-19 from Anthropic official pricing. All prices in USD per 1M tokens. See methodology and price changelog.