Opus 4.7 Costs More Than 4.6 Despite Flat Pricing
Anthropic's new Opus 4.7 model has a new tokenizer that breaks text into up to 47% more tokens, resulting in significantly higher costs for users despite the per-token price remaining the same as the previous version.
Why it matters
This change in Opus 4.7 has significant cost implications for Anthropic's customers, potentially impacting their AI development and deployment budgets.
Key Points
- 1Opus 4.7 has a new tokenizer that generates up to 47% more tokens for the same text
- 2This leads to significantly higher costs for users, despite Anthropic maintaining the same per-token pricing
- 3Early measurements show the impact of this change on costs for Claude Code users
Details
Anthropic's latest Opus 4.7 model maintains the same per-token pricing as the previous 4.6 version. However, the introduction of a new tokenizer in 4.7 results in the same text being broken down into up to 47% more tokens. This means that despite the flat pricing, each request ends up costing significantly more for users of the Claude Code platform. The article provides early measurements demonstrating the practical impact of this change, highlighting how the increased token counts translate to higher overall costs for customers despite Anthropic's efforts to keep the per-token price consistent.
No comments yet
Be the first to comment