Claude Opus 4.7 API pricing: $15 per million input tokens, $75 per million output. When the most-capable tier earns its premium.
Claude Opus 4.7 sits at the top of the 2026 lineup. It is the most capable model Anthropic offers, and the most expensive — 5× over Sonnet on both input and output.
What is Claude Opus 4.7 priced at?
Claude Opus 4.7 costs $15 per million input tokens and $75 per million output tokens. Cache write is $18.75/M and cache reads are $1.50/M. Batch API halves all rates: $7.50 input, $37.50 output.
When is Claude Opus 4.7 worth the premium over Sonnet?
When the task is multi-step agentic (where errors compound), requires multi-hop reasoning over 100k+ token contexts, or is high-stakes (legal, medical, financial). For templated tasks with a fixed schema, Sonnet matches on quality at 1/5 the price.
Is Claude Opus 4.7 the most capable AI model available?
As of 2026, Opus 4.7 is Anthropic's most capable model. It leads on multi-step reasoning, long-context retrieval, and agentic task completion benchmarks. Head-to-head with GPT-4 and Gemini Ultra, results vary by task type — no single model dominates all categories.