Opus 4.7 uses more tokens than Opus 4.6 at identical per-token pricing
Simon Willison's token counter data shows Opus 4.7 consuming 1.46x more tokens for text and up to 3x more for high-res images vs Opus 4.6.
Anthropic's Claude Opus 4.7 consumes measurably more tokens than its predecessor at the same per-token price, according to analysis by Simon Willison. Willison upgraded his Claude token counter tool to compare models side-by-side and found Opus 4.7 using roughly 1.46x more tokens for raw text and up to 3x more for images compared to Opus 4.6. Since both models carry identical per-token pricing, the higher token consumption translates directly into higher costs for the same workload.
Willison's testing adds nuance: a 30-page, text-heavy PDF came in at 1.08x the token count rather than 1.46x, suggesting the multiplier varies with content type. The image gap closes significantly for lower-resolution inputs — a 682×318 pixel image cost 314 tokens on Opus 4.7 versus 310 on Opus 4.6, nearly identical. The large image multiplier appears tied to Opus 4.7's ability to process higher-resolution inputs, not an across-the-board increase. Developers and teams running cost models based on Opus 4.6 benchmarks should re-evaluate expected spend, particularly for text-heavy or high-resolution image workloads before migrating.