Anthropic Claude API Pricing 2026: Sonnet 4, Haiku, Opus Costs
Per-million-token pricing for the current Claude API lineup — Sonnet 4, Haiku 3.5 and Opus 4. Anthropic's prompt caching cuts input cost by up to 90%, making Claude surprisingly competitive on conversation-heavy workloads. Cross-compare on the main calculator.
Claude caching saves up to 90% on input
Claude Haiku 3.5 cached input is $0.08/M versus $0.80/M standard — a 90%discount. That is far deeper than OpenAI's 50% cache cap.
Cheapest Claude model for coding
Claude Haiku 3.5 — $0.80 input / $4.00 output per million tokens.
Anthropic pricing per million tokens
USD per 1,000,000 tokens. Cached rate applies to repeated prompt prefixes; Batch rate requires async (up to 24-hour) turnaround.
| Model | Input $/M | Output $/M | Cached $/M | Batch in $/M | Batch out $/M | Context |
|---|---|---|---|---|---|---|
| Claude Haiku 3.5 | $0.80 | $4.00 | $0.08 | $0.40 | $2.00 | 200K |
| Claude Sonnet 4 | $3.00 | $15.00 | $0.30 | $1.50 | $7.50 | 200K |
| Claude Opus 4 | $15.00 | $75.00 | $1.50 | $7.50 | $37.50 | 200K |
Claude Haiku 3.5
$0.80 in / $4.00 out · 200K context
Cache saves 90% on input
Claude Sonnet 4
$3.00 in / $15.00 out · 200K context
Cache saves 90% on input
Claude Opus 4
$15.00 in / $75.00 out · 200K context
Cache saves 90% on input
When Claude wins on cost
- Repeat-heavy prompts:Anthropic's 90% cache discount beats OpenAI's 50% cap whenever long prefixes recur.
- Long-context tasks: every Claude model here ships a 200K context window.
- High-volume chat: Haiku 3.5 at $0.80/$4.00 is purpose-built for support-style traffic.
Compare Claude against other providers
Frequently asked questions
How much does Claude Sonnet 4 cost?
Claude Sonnet 4 lists at $3.00 per million input tokens and $15.00 per million output tokens. With prompt caching, repeated input drops to $0.30/M — a 90% discount on cached tokens. The Batch API halves both rates to $1.50/M input and $7.50/M output for async workloads.
Claude vs GPT-4o cost — which is cheaper?
On list pricing, Claude Sonnet 4 ($3/$15) is more expensive than GPT-4o ($2.50/$10) per million tokens. However, Anthropic's prompt caching is far more aggressive: cached Sonnet 4 input is $0.30/M versus GPT-4o's $1.25/M. For conversation-heavy or long-context-repeat workloads, Claude often wins on effective cost. For one-shot generation, GPT-4o is typically cheaper.
How much does Claude Opus 4 cost?
Claude Opus 4 is Anthropic's premium reasoning tier at $15/M input and $75/M output — the most expensive model in the lineup. Prompt caching brings cached input down to $1.50/M (90% off). Reserve Opus 4 for the hardest reasoning tasks; Sonnet 4 handles most production workloads at one-fifth the price.
Is Claude Haiku 3.5 cheap enough for high-volume chat?
Yes. Claude Haiku 3.5 at $0.80/M input and $4.00/M output is the cheapest Claude model, and cached input drops to $0.08/M — 90% off. For high-volume support bots, classification and short-form chat, Haiku 3.5 is typically the right Claude pick.
How does Anthropic prompt caching compare to OpenAI's?
Anthropic discounts cached input by up to 90% (e.g. Claude Haiku 3.5: $0.80 → $0.08/M). OpenAI caps cached input at 50% off. For workloads that repeat long prompts, Anthropic's deeper cache discount often produces the lower effective bill despite a higher list price.