Open Calculator →

Claude Haiku Pricing:
Cost Per Token Calculator 2025

Claude Haiku 3.5 is Anthropic's fastest, most affordable model. Estimate your exact API costs — or paste your Claude Code session log for a precise breakdown.

Calculate My Costs →

Claude Haiku 3.5 — Official Pricing (2025)

Token Type Price per 1M Tokens Notes
Input (standard) $0.80 Fresh context on every call
Output $4.00 Generated tokens
Cache write $1.00 First call that caches a block
Cache read Best value $0.08 Subsequent calls hit cache — 10× cheaper

What Does Claude Haiku Cost in Practice?

At $0.80/M input and $4/M output, Claude Haiku is optimized for high-volume, latency-sensitive tasks. Here are ballpark costs for common workloads:

Frequently Asked Questions

How much does Claude Haiku cost per token in 2025?

Claude Haiku 3.5 costs $0.80 per million input tokens and $4.00 per million output tokens. Prompt cache writes are $1.00/M and cache reads are $0.08/M — making cached prompts up to 90% cheaper on repeated calls.

Is Claude Haiku cheaper than GPT-4o mini?

Claude Haiku and GPT-4o mini are comparably priced for most workloads. Haiku's prompt caching gives it a strong advantage at scale: cached context costs only $0.08/M — roughly 10× less than a fresh read. For sessions with long, repeated system prompts, Haiku caching typically saves 40–70% of total bill.

When should I use Haiku instead of Claude Sonnet?

Use Haiku for: high-volume routing or classification tasks, lightweight sub-agents in a multi-agent pipeline, or any loop where speed and cost matter more than reasoning depth. Switch to Claude Sonnet 4.5 for complex code generation, deep document analysis, or tasks that regularly need extended thinking.

How do I calculate my exact Claude Haiku costs from Claude Code session logs?

Claude Code writes session logs to ~/.claude/projects/<project>/*.jsonl. Each line records the model, token counts (input, output, cache write, cache read), and timestamp. Paste the file into the Claude Code Cost Calculator above — it applies per-model pricing and gives a breakdown by model, tool, and hour. Runs entirely in your browser; nothing is stored.