Claude Pricing 2026:
What Each Claude Model Costs to Run
A plain-English overview of what Anthropic's Claude models cost in 2026 — Haiku 4.5, Sonnet 5.5, Opus 5.5 and Fable 5.1 — with worked monthly cost examples and how Claude compares with OpenAI. Last verified: 2026-10-01.
Claude is billed per token through Anthropic's API: you pay for the tokens you send (input) and the tokens Claude writes (output). Output costs five times as much as input on every current model. For the full rate card — prompt caching, batch prices and previous-generation models — see our Claude API pricing guide.
Which Claude Model Should You Pay For?
| Use Case | Best Model | Why |
|---|---|---|
| High-volume chatbots, classification | Haiku 4.5 | Half the price of Sonnet 5.5, fast, handles most routine tasks |
| Customer support, summarization | Sonnet 5.5 | Best quality/cost ratio for complex tasks |
| Complex coding, analysis, research | Sonnet 5.5 | Production default for demanding work, 1M context |
| Long agentic tasks, hard analysis | Opus 5.5 | Higher accuracy at twice Sonnet's price |
| Hardest problems where quality outweighs cost | Fable 5.1 | Anthropic's most capable widely released model |
Claude vs OpenAI Pricing Comparison
| Tier | Claude | OpenAI | Input / output per 1M |
|---|---|---|---|
| Budget | Haiku 4.5 | GPT-6 Luna | $1/$5 vs $0.10/$0.50 — OpenAI is 10× cheaper |
| Balanced | Sonnet 5.5 | GPT-6.1 Sol | $2/$10 vs $2/$10 — identical |
| Premium | Opus 5.5 | GPT-6 Astra | $4/$20 vs $10/$50 — Claude is 2.5× cheaper |
| Top tier | Fable 5.1 | GPT-6 Astra | $10/$50 vs $10/$50 — identical |
Prompt Caching Cuts Repeated Input
When the same prompt prefix is reused — a long system prompt, a shared document, a tool list — Claude bills it at the cache-read rate instead of the full input price:
- Sonnet 5.5: $2.00/M → cache read $0.20/M (90% off)
- Opus 5.5: $4.00/M → cache read $0.20/M (95% off)
- Haiku 4.5: $1.00/M → cache read $0.10/M (90% off)
Best for: RAG systems with large context, long system prompts, and multi-turn conversations with a fixed system prompt.
Real-World Claude Cost Examples
Customer Support Bot (50,000 conversations/month)
- System prompt: 2,000 tokens (cached after the first call)
- Per conversation: 500 tokens fresh input + 300 tokens output
- Without caching (Sonnet 5.5): $2 × 125M + $10 × 15M = $250 + $150 = $400/month
- With caching: $0.20 × 100M + $2 × 25M + $150 output = ~$220/month
- On Haiku 4.5 with caching: $10 + $25 + $75 = ~$110/month
Document Analysis Pipeline (10,000 docs/month)
- Average: 5,000 tokens input + 800 tokens output
- Haiku 4.5: $50 + $40 = $90/month
- Sonnet 5.5: $100 + $80 = $180/month
- Opus 5.5: $200 + $160 = $360/month
Claude.ai Subscriptions vs the API
The Claude.ai apps are sold as monthly subscriptions, separate from API billing. If you are building a product, you need the API and pay per token; a Claude.ai subscription does not include API usage. For a side-by-side of consumer AI plans, see the AI subscription comparison.
Calculate Your Claude API Costs
Compare Claude Haiku vs Sonnet vs GPT-6 for your specific usage.
AI Cost Calculator