Claude Opus 4.7 is priced at $5/M input tokens and $25/M output tokens — the same sticker price as Opus 4.6. The catch: Opus 4.7 ships with a new tokenizer that produces 1.0–1.46× more tokens for the same English text. For code and JSON, the multiplier hits the upper end. At identical per-token prices, that's a quiet 20–46% price hike for anyone migrating up from Opus 4.6.
This calculator models the real cost. Paste your actual prompt — system message, RAG context, the works — and we'll tokenize it through Anthropic's official count_tokens API (no character approximations). Toggle prompt caching to see the 90% input discount applied to cached portions. Toggle the Batch API for the 50% async discount. Compare side-by-side against Opus 4.6, Sonnet 4.6, and Haiku 4.5 at the bottom.
All numbers verified against claude.com/pricing on the date shown. Code is open source. No accounts, no tracking, no ads.
Frequently asked questions
How much does Claude Opus 4.7 cost?
$5 per million input tokens, $25 per million output tokens. With prompt caching, cached input drops to $0.50 per million.
Is Opus 4.7 more expensive than Opus 4.6?
The sticker prices are identical. But Opus 4.7's tokenizer produces ~20% more tokens for prose and up to 46% more for code, meaning the same English prompt costs 20–46% more on 4.7.
Does Opus 4.7 support prompt caching?
Yes. Cache writes cost 1.25× input rate (5-minute) or 2× (1-hour). Cache reads cost 10% of input rate.
What context window does Opus 4.7 support?
1 million tokens at flat pricing — no tier jump above 200K, unlike Gemini.
When is Opus 4.7 worth the cost over Sonnet 4.6?
For tasks where you've benchmarked and measured Sonnet failing. For routine work, Sonnet 4.6 at $3/$15 lands within 1% of GPT-5.4's bill.