Opus 4.7 Tokenizer

Description

New tokenizer shipped with Opus 4.7 that produces 1.0-1.35x more tokens than Opus 4.6 for identical text. English-heavy prose sits near ~1.1x; code-heavy prompts with long identifiers, generics, and nested constructs can hit ~1.35x. Per-token pricing unchanged, so worst-case effective cost rises ~35%. Interacts with prompt caching: cache-friendly workflows can partially absorb the overhead; cache-hostile workflows just pay more.

Key claims

Relations

Sources

src-20260423-22e662f6932a