Simon Willison upgrades Claude Token Counter with model comparison feature
Simon Willison has made an upgrade to his Claude Token Counter tool, which now allows for better comparison of token counts across different AI models. This update focuses on the differences between Claude Opus versions 4.7 and 4.6, mainly due to a new tokenizer in 4.7 that changes how text is processed. According to Anthropic, this tokenizer can increase token counts by 1.0 to 1.35 times based on the type of content. Tests showed Opus 4.7 used 1.46 times more tokens than 4.6 for the same prompts. Both models cost $5 per million input tokens and $25 for output, but Opus 4.7 ends up being about 40% more expensive. The tool also supports higher-resolution images, with specific token count increases for various image sizes. Willison announced this on April 20, 2026.
Key facts
- Simon Willison upgraded the Claude Token Counter tool to compare token counts across different models
- Opus 4.7 uses an updated tokenizer that can increase token counts by 1.0-1.35× depending on content
- Opus 4.7 used 1.46× more tokens than Opus 4.6 for the same system prompt
- Opus 4.7 pricing remains $5 per million input tokens and $25 per million output tokens
- Token inflation makes Opus 4.7 approximately 40% more expensive than Opus 4.6
- Opus 4.7 supports images up to 2,576 pixels on the long edge
- A 3456x2234 pixel PNG image showed a 3.01× token increase for Opus 4.7 compared to 4.6
- A 15MB text-heavy PDF resulted in a 1.08× token multiplier for Opus 4.7
Entities
Artists
- Simon Willison
Institutions
- Anthropic