Claude Opus 4.7 uses up to 3x more tokens than Opus 4.6
@simonw upgraded a Claude token counter tool and found that Opus 4.7 uses 1.46x more tokens for text and up to 3x more tokens for images compared to Opus 4.6, making it effectively more expensive despite identical per-token pricing.
Score breakdown
Developers budgeting for Claude API usage — especially image-heavy pipelines — should re-benchmark their token costs when migrating from Opus 4.6 to Opus 4.7, as real-world spend could be significantly higher than per-token pricing suggests.
- 01@simonw upgraded a Claude token counter tool to compare token usage across different models.
- 02Opus 4.7 uses 1.46x more tokens than Opus 4.6 for text inputs.
- 03Opus 4.7 uses up to 3x more tokens than Opus 4.6 for image inputs.
@simonw upgraded a Claude token counter tool to enable side-by-side model comparisons and used it to measure token consumption differences between Opus 4.7 and Opus 4.6. The findings show that Opus 4.7 uses 1.46x more tokens for text inputs and up to 3x more tokens for image inputs relative to Opus 4.6.
Because Anthropic prices Opus 4.7 at the same per-token rate as Opus 4.6, the higher token counts effectively represent a significant price increase for real-world workloads — particularly for image-heavy use cases where costs could triple.
Key facts
- 01@simonw upgraded a Claude token counter tool to compare token usage across different models.
- 02Opus 4.7 uses 1.46x more tokens than Opus 4.6 for text inputs.
- 03Opus 4.7 uses up to 3x more tokens than Opus 4.6 for image inputs.
- 04Opus 4.7 is priced the same as Opus 4.6 on a per-token basis.
- 05The higher token consumption makes Opus 4.7 effectively more expensive despite identical listed pricing.
Topics
Summary and scoring are generated automatically from the original article. We always link back to the publisher and never republish images or paywalled content. Last processed Apr 21, 2026 · 18:16 UTC. How this works →